diff --git a/checkpoint-10000/optimizer.bin b/checkpoint-10000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..90a56b6ce4c1db03ea26fdcbf61e47a1e91a4435
--- /dev/null
+++ b/checkpoint-10000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:825fbb3256e9bcbf5a11267139e74509291cbab7f436320b28037ada1330383d
+size 260158662
diff --git a/checkpoint-10000/random_states_0.pkl b/checkpoint-10000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..dfa06933588b7465e205a742b0fc7c56aa846d69
--- /dev/null
+++ b/checkpoint-10000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d043cdc24ce0e0a382e1f82d56ede70aebb4993290f6e65e72b7352e84688bd
+size 16100
diff --git a/checkpoint-10000/scheduler.bin b/checkpoint-10000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..c569fedaa4e85fe14bb34d78fec4d8f4e79f66eb
--- /dev/null
+++ b/checkpoint-10000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e029419f0f150eb1d92bf9ccc8c0031ba9e70dd9452332f5836ed93bec14f72
+size 1000
diff --git a/checkpoint-10000/transformer/config.json b/checkpoint-10000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-10000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-10000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-10000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-10000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-10000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-10000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-10000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-10000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-10000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-10000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-10000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-10000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-10000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-10000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-10000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..aefa4126dd8e1f5c769a183e46edc959ebf18e34
--- /dev/null
+++ b/checkpoint-10000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f289eedc1830b7454d2939e32ac4bacd6597562cfed0a00ad24fd9662676ee08
+size 1702614662
diff --git a/checkpoint-10000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-10000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-10000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-15000/optimizer.bin b/checkpoint-15000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..5efbc903c916d9456071d03c0c6a4875d2c7b9eb
--- /dev/null
+++ b/checkpoint-15000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e3815f51fd8b00abb4fff483f13c01aff6bc0f2327e06540e587098945adda7
+size 260158662
diff --git a/checkpoint-15000/random_states_0.pkl b/checkpoint-15000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..db9d9a2ad2e0873c6743025e1c75ecbd5b225a92
--- /dev/null
+++ b/checkpoint-15000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a0f46b15d56a6d7bf9b03751bf2ef90166d92e3268c7a38d7f0e6744876f95f
+size 16100
diff --git a/checkpoint-15000/scheduler.bin b/checkpoint-15000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..ec88222391cc6b8ee8e195fe89328c4c273195c8
--- /dev/null
+++ b/checkpoint-15000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8e5e28854899a30817e7e3d629246143556db2c6cc66aa1f102c5264991da1f
+size 1000
diff --git a/checkpoint-15000/transformer/config.json b/checkpoint-15000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-15000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-15000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-15000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-15000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-15000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-15000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-15000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-15000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-15000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-15000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-15000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-15000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-15000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-15000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-15000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..83ef38d539a08b652cda6fd466c1595c21b27d9b
--- /dev/null
+++ b/checkpoint-15000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0c19016e7cccb48a26542c7a1c8713e7f0ae4dfd2cff3d80c5e1f5aa7b963a3
+size 1702614662
diff --git a/checkpoint-15000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-15000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-15000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-20000/optimizer.bin b/checkpoint-20000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..47f5cf6629af72ebd7acfdaafd9af9acf8c85959
--- /dev/null
+++ b/checkpoint-20000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c6a16be10e965eb483ddbf9402ff7b412cb2f41a624bd69ac16d09912fd2514
+size 260158662
diff --git a/checkpoint-20000/random_states_0.pkl b/checkpoint-20000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c3e28ddc3979913766e4ce165ba04c43ef950dc5
--- /dev/null
+++ b/checkpoint-20000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6124b4614d022548c4e0c4f2666dacef2405bded7b9e89869fc4d0ef191b7e8c
+size 16036
diff --git a/checkpoint-20000/scheduler.bin b/checkpoint-20000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..9132cb9311b46010a27c7f03aebe78918a79bfc0
--- /dev/null
+++ b/checkpoint-20000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20d3e40b3de445f613654954d5709bf0c8df590da0130398a3c202dd084fa442
+size 1000
diff --git a/checkpoint-20000/transformer/config.json b/checkpoint-20000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-20000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-20000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-20000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-20000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-20000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-20000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-20000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-20000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-20000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-20000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-20000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-20000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-20000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-20000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-20000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..352e1446734368593c44d251553ea3fa0ae61517
--- /dev/null
+++ b/checkpoint-20000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a601394d588fae91374d9ac278c421580563c22e7dc58aade55bbf678b3d97f1
+size 1702614662
diff --git a/checkpoint-20000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-20000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-20000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-25000/optimizer.bin b/checkpoint-25000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..35a98fd80a029294bfbad87f19f6b845af9281fb
--- /dev/null
+++ b/checkpoint-25000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a75ed0a1bd8b73dd72e5d8f7fbe5ad102fc1cd2f76bebd831d58a3b71940aa0
+size 260158662
diff --git a/checkpoint-25000/random_states_0.pkl b/checkpoint-25000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..d6a425553c555c7b168ceef7b9fc025bf8817e43
--- /dev/null
+++ b/checkpoint-25000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10072ff99780ddb637d9349e65165381233f4075e05c5a76c989ef4058e94063
+size 16100
diff --git a/checkpoint-25000/scheduler.bin b/checkpoint-25000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0bc3a9f3549db053353f686cd24ac3fb2a3a840d
--- /dev/null
+++ b/checkpoint-25000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84b9c577c98c6c74cf851e30205dcf7e7e5b7b7dc5fd38753a0e3ac92bda8bca
+size 1000
diff --git a/checkpoint-25000/transformer/config.json b/checkpoint-25000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-25000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-25000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-25000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-25000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-25000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-25000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-25000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-25000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-25000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-25000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-25000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-25000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-25000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-25000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-25000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..52a5f5198ad59224ad948293226959f5ad6e7b4f
--- /dev/null
+++ b/checkpoint-25000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:773fff6e1d64bc08f26e8d8938084421df85cfbbd008390d454d5b4db0e753fa
+size 1702614662
diff --git a/checkpoint-25000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-25000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-25000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-30000/optimizer.bin b/checkpoint-30000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0a682359a1f0b5569142bf1c0ed8897144207012
--- /dev/null
+++ b/checkpoint-30000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7efea8149f9c086e1f85ab3852efc30176bb9324e38e30e397e8c2b630d95b6
+size 260158662
diff --git a/checkpoint-30000/random_states_0.pkl b/checkpoint-30000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..08b2d78e3f53d72728338a04042d42feb1662798
--- /dev/null
+++ b/checkpoint-30000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1607fb2165f4678566a9ce2001b0baa4e4523266f27be7cf24a587f5f604a60
+size 16036
diff --git a/checkpoint-30000/scheduler.bin b/checkpoint-30000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..ca76f60fb1e774fa80f727de4708dc5800f19eb0
--- /dev/null
+++ b/checkpoint-30000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c165903743ec11fd70e4c46a4080ac178b4d10421403edb6d00fa61631fdf1e2
+size 1000
diff --git a/checkpoint-30000/transformer/config.json b/checkpoint-30000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-30000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-30000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-30000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-30000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-30000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-30000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-30000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-30000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-30000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-30000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-30000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-30000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-30000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-30000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-30000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..9fb638013a4fe8c1d825d4c672ddfed98b4f0aa4
--- /dev/null
+++ b/checkpoint-30000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13a1df4a4423f083cc612f4740d3c8ac24667552d3bbf41d3d22d60ccac3ca44
+size 1702614662
diff --git a/checkpoint-30000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-30000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-30000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-35000/optimizer.bin b/checkpoint-35000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..98a15e737648cba2ed6b06513147de3863745cc1
--- /dev/null
+++ b/checkpoint-35000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d9c14fd81b2d2e2d3d01c2e2c686bb28f1686ebc7d41f3e265470ef18a75549
+size 260158662
diff --git a/checkpoint-35000/random_states_0.pkl b/checkpoint-35000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..4f415f02b68249d13b0820bba4f544f6f90bfabb
--- /dev/null
+++ b/checkpoint-35000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71441866abcd0ade9287f3135e3778d56c4b9a2a4842c1506ada1e2dfa7c44f9
+size 16036
diff --git a/checkpoint-35000/scheduler.bin b/checkpoint-35000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..f349bcdf345bae56f4f36daebc5b4416d4948216
--- /dev/null
+++ b/checkpoint-35000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:775096dff0cdcbbfcfa938e415c309928b0f80376462ad6939aa1fab335e37be
+size 1000
diff --git a/checkpoint-35000/transformer/config.json b/checkpoint-35000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-35000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-35000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-35000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-35000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-35000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-35000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-35000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-35000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-35000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-35000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-35000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-35000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-35000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-35000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-35000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..17fbb83a9b822c51bc568aff442a216fdfe2af9b
--- /dev/null
+++ b/checkpoint-35000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3493f9f0b2ca7213f086018e0e80fab0742f3802fedefb8b1594a3bd86664f1
+size 1702614662
diff --git a/checkpoint-35000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-35000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-35000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-40000/optimizer.bin b/checkpoint-40000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..8aa018362994bbb4e95305d2f8049d82429797ae
--- /dev/null
+++ b/checkpoint-40000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eb8698dc7032455c5003456b6f8f77e69356c347bd1a9a091735a3f49f0cd62
+size 260158662
diff --git a/checkpoint-40000/random_states_0.pkl b/checkpoint-40000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e3f7a6554608accc9d2d167348ef5c2ee8ba00ae
--- /dev/null
+++ b/checkpoint-40000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:098c8e5589cb1f75eddf76f99e4f73c7bac87f211656f31b5004de06ac1d13ca
+size 16100
diff --git a/checkpoint-40000/scheduler.bin b/checkpoint-40000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..3d1ff462f275c5e2ed3e9ed2a071ff0933b960a9
--- /dev/null
+++ b/checkpoint-40000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d81566d62adf29a46890d9e6d98ef1314cccae7aa70a33babc6e0118914db05
+size 1000
diff --git a/checkpoint-40000/transformer/config.json b/checkpoint-40000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-40000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-40000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-40000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-40000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-40000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-40000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-40000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-40000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-40000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-40000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-40000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-40000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-40000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-40000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-40000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..f22a8dc3c1f8f4f1a8d0ed964d3151d368ec6ca9
--- /dev/null
+++ b/checkpoint-40000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b724a6364e135fad67881df7c90d94a3c3b192c7aa2cc6ebbd8f22e17108e18
+size 1702614662
diff --git a/checkpoint-40000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-40000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-40000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-45000/optimizer.bin b/checkpoint-45000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..83a9bd65ef8116c7478fa0449b9c22dbdf1a8f37
--- /dev/null
+++ b/checkpoint-45000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b48f56bc2124728b59aaae2c4992529e6b9101d942f813155c5f8623e3dc75bb
+size 260158662
diff --git a/checkpoint-45000/random_states_0.pkl b/checkpoint-45000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..6304f6436583bb5d9efd695c2dd64e5812d75831
--- /dev/null
+++ b/checkpoint-45000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d235cd3b740703f9ebba4dd29d920929979fe0ac9466392fee0add771d01dd2
+size 16100
diff --git a/checkpoint-45000/scheduler.bin b/checkpoint-45000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..169c502a12b78ce9d16e7e5960f908ccb924e303
--- /dev/null
+++ b/checkpoint-45000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cbf184d99383256b9dded2287dc286184e6188d4fbdb1e9a5c24f00ebf6f6cd
+size 1000
diff --git a/checkpoint-45000/transformer/config.json b/checkpoint-45000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-45000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-45000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-45000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-45000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-45000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-45000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-45000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-45000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-45000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-45000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-45000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-45000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-45000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-45000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-45000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..53dc3bccd41addf0df25ee060758a36951a09e59
--- /dev/null
+++ b/checkpoint-45000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2e1f00b8691857a6a371a9905267f59bcc247aeff7605295fb3b267b45aa980
+size 1702614662
diff --git a/checkpoint-45000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-45000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-45000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-5000/optimizer.bin b/checkpoint-5000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fa209409ffd08df406f558bc6075241febc6a27d
--- /dev/null
+++ b/checkpoint-5000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ed92001eabfb75c17bb0803e9a1d386036443ab7790b29aa6e7314158bb8e23
+size 260158662
diff --git a/checkpoint-5000/random_states_0.pkl b/checkpoint-5000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..7b2bdf59605a5ae3c8ac8c3063290f83573ba11e
--- /dev/null
+++ b/checkpoint-5000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee9b1ceeaf1fb2aba4e4ee5d2822e379294a221066a516f33ed30005a4a7aba3
+size 16036
diff --git a/checkpoint-5000/scheduler.bin b/checkpoint-5000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..f288b4add0d545c2a09eaef5eb6d976959835de2
--- /dev/null
+++ b/checkpoint-5000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5a66b43b1f48b77b76fc9e12b8669e23ca436380930f17d7762d11aeb1b7b06
+size 1000
diff --git a/checkpoint-5000/transformer/config.json b/checkpoint-5000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-5000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-5000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-5000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-5000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-5000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-5000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-5000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-5000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-5000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-5000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-5000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-5000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-5000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-5000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-5000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..25c15308f4920bf7ad2adaf5e16e8f9137f12085
--- /dev/null
+++ b/checkpoint-5000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3524ebd1f9e7a593cece97d2b890a3fe9eccf87b3d5ac2dbee9d638a400eed7
+size 1702614662
diff --git a/checkpoint-5000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-5000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-5000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-50000/optimizer.bin b/checkpoint-50000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..454896733b475aedd63219e24bf163696dfa32ed
--- /dev/null
+++ b/checkpoint-50000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:252b021df7d8a88e3737bb18a564824ffd1a9441bde0609407387d64bd7c11f9
+size 260158662
diff --git a/checkpoint-50000/random_states_0.pkl b/checkpoint-50000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..cb6a2579579e73a5ff95a053f3272f3e506fa3d1
--- /dev/null
+++ b/checkpoint-50000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a644c95d6d5c29036b93bf75f56ea0784f5d0df73b909a494fa8639404ace626
+size 16100
diff --git a/checkpoint-50000/scheduler.bin b/checkpoint-50000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..d1047713ce179180fd8d61ca66710fde243fec3e
--- /dev/null
+++ b/checkpoint-50000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73a5704ffb97c9a254733ef57110528308bbecb45d1174a1487cc05d2a061cb0
+size 1000
diff --git a/checkpoint-50000/transformer/config.json b/checkpoint-50000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-50000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-50000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-50000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-50000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-50000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-50000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-50000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-50000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-50000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-50000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-50000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-50000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-50000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-50000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-50000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..4f11ca815ce3482001c78f6a4b1cbb11e61ff502
--- /dev/null
+++ b/checkpoint-50000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:876716dda96f169bb2717eccd80848bd95dfaf52913fe08475acb56c48be8036
+size 1702614662
diff --git a/checkpoint-50000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-50000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-50000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-55000/optimizer.bin b/checkpoint-55000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..834a145e9635ad035129fa4f40cd94aab174dbed
--- /dev/null
+++ b/checkpoint-55000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fd05044a3c65d341d8cdc795ce991dcccc322b6f80a5b3bd0974136751980a9
+size 260158662
diff --git a/checkpoint-55000/random_states_0.pkl b/checkpoint-55000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..20cfbecfed6ad7eaaa8ba5639f489dbe8fcb25b8
--- /dev/null
+++ b/checkpoint-55000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d50b39e771d891923eb5f64a1107519761691e65468d953c23a9c471d517c507
+size 16100
diff --git a/checkpoint-55000/scheduler.bin b/checkpoint-55000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..3f4fd31a28f443b8aabf4539e792257d15f04f10
--- /dev/null
+++ b/checkpoint-55000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58dcd2b6d146be1443f2b9ad98c64cdb40333f1f5cddc9409175236eb874ebe4
+size 1000
diff --git a/checkpoint-55000/transformer/config.json b/checkpoint-55000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-55000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-55000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-55000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-55000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-55000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-55000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-55000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-55000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-55000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-55000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-55000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-55000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-55000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-55000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-55000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..c76b2245dc53b78bd48ee3657578ce543ceeae69
--- /dev/null
+++ b/checkpoint-55000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3bec1fc9c262a5dd83ea56114c7f57e63512079749431858ab3fd1d675bb7eb
+size 1702614662
diff --git a/checkpoint-55000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-55000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-55000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-60000/optimizer.bin b/checkpoint-60000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..5f3c373b4658c019ea77f388b69c57b6c5147633
--- /dev/null
+++ b/checkpoint-60000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46c85e87edb2285660c516630fb1a564f08e5e5c44f71e2cca336c5e13d90e6a
+size 260158662
diff --git a/checkpoint-60000/random_states_0.pkl b/checkpoint-60000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..0675959629dbd5e9d93b1d8b025b83dcb7b44276
--- /dev/null
+++ b/checkpoint-60000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e4a7a531b71423266c76c813dc306c44c385c12e457e3eda0d5787ff9a051f0
+size 16036
diff --git a/checkpoint-60000/scheduler.bin b/checkpoint-60000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..3095111002a26f34417fdddecdd4bbc21d6a8cfa
--- /dev/null
+++ b/checkpoint-60000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8519e97f48807e8442183da8f5fc247656fed6b7f3fd4336528c3c9d250ce2f2
+size 1000
diff --git a/checkpoint-60000/transformer/config.json b/checkpoint-60000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-60000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-60000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-60000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-60000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-60000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-60000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-60000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-60000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-60000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-60000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-60000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-60000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-60000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-60000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-60000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..f2c5907e0ffdfe75eeda36bd3b12372b877f17d9
--- /dev/null
+++ b/checkpoint-60000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7427ffeb3b2de8082fb8d3854caf9357dff66d6beae7af0d6c2566e55907a66e
+size 1702614662
diff --git a/checkpoint-60000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-60000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-60000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}
diff --git a/checkpoint-65000/optimizer.bin b/checkpoint-65000/optimizer.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0f6886756c94bb6082bfde73613bac62d5fad84d
--- /dev/null
+++ b/checkpoint-65000/optimizer.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7dc26c1abdd76bbca21f05363b26f94d7618607b5552f66ae74b17fd7d2d3ace
+size 260158662
diff --git a/checkpoint-65000/random_states_0.pkl b/checkpoint-65000/random_states_0.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..273ee4e354258a4062d863fe855d3bed581a48ce
--- /dev/null
+++ b/checkpoint-65000/random_states_0.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22a37f37a3eba47b7c60df0fa32e18f4a1e0df9aa9b00473607257a505123560
+size 16036
diff --git a/checkpoint-65000/scheduler.bin b/checkpoint-65000/scheduler.bin
new file mode 100644
index 0000000000000000000000000000000000000000..102ec551270338fa0ed976077ca0422cf359b3ea
--- /dev/null
+++ b/checkpoint-65000/scheduler.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab997fe0d21024e70d4ac3bf4643f58bbd0bf75e3fa9afd8975a4174cdd52e78
+size 1000
diff --git a/checkpoint-65000/transformer/config.json b/checkpoint-65000/transformer/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..c98351eb2b27f34e8081200d6387ea7c9cd0493d
--- /dev/null
+++ b/checkpoint-65000/transformer/config.json
@@ -0,0 +1,4 @@
+{
+  "_class_name": "QwenVLSD3_DirectMap_Transformer2DModel",
+  "_diffusers_version": "0.32.0.dev0"
+}
diff --git a/checkpoint-65000/transformer/diffusion_pytorch_model-00001-of-00005.bin b/checkpoint-65000/transformer/diffusion_pytorch_model-00001-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..0b79b5e93a929bae1bd48bf9b566596b7d84f02b
--- /dev/null
+++ b/checkpoint-65000/transformer/diffusion_pytorch_model-00001-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07e9802d73a359782f33e21bfff3ca94a068735f3eac23e8902326352f72293
+size 9967776816
diff --git a/checkpoint-65000/transformer/diffusion_pytorch_model-00002-of-00005.bin b/checkpoint-65000/transformer/diffusion_pytorch_model-00002-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e91c78d5ca28d50ce243a8c2b132b1725c5df508
--- /dev/null
+++ b/checkpoint-65000/transformer/diffusion_pytorch_model-00002-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff6bff969fe794615f8fb25ec4e4b4f04000040523880ddc8a613f275785843
+size 9987471140
diff --git a/checkpoint-65000/transformer/diffusion_pytorch_model-00003-of-00005.bin b/checkpoint-65000/transformer/diffusion_pytorch_model-00003-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..22515e727458a4ed7be086d13f8ff401c4872989
--- /dev/null
+++ b/checkpoint-65000/transformer/diffusion_pytorch_model-00003-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5d03430f675c4c44930f75514554e6726c1693b23d84522c57e873bb84f0ee9
+size 9940103750
diff --git a/checkpoint-65000/transformer/diffusion_pytorch_model-00004-of-00005.bin b/checkpoint-65000/transformer/diffusion_pytorch_model-00004-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..fb00347ba032e0159e6598c9bc657f958b38d227
--- /dev/null
+++ b/checkpoint-65000/transformer/diffusion_pytorch_model-00004-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef4c53fb68f21cb007c78d1e45b0f9742d1897bbd614a2ddc0da31af1d15add
+size 9987474226
diff --git a/checkpoint-65000/transformer/diffusion_pytorch_model-00005-of-00005.bin b/checkpoint-65000/transformer/diffusion_pytorch_model-00005-of-00005.bin
new file mode 100644
index 0000000000000000000000000000000000000000..94dcab56335c1be00dea62c0b725732f725655d0
--- /dev/null
+++ b/checkpoint-65000/transformer/diffusion_pytorch_model-00005-of-00005.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b5be73dbd7e243d054b7ee90a9a88a9a17d5a50c46d26e3a860fe81aba6dceb
+size 1702614662
diff --git a/checkpoint-65000/transformer/diffusion_pytorch_model.bin.index.json b/checkpoint-65000/transformer/diffusion_pytorch_model.bin.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a4e4519988063eda3484cd1a319bdbf3279c4f
--- /dev/null
+++ b/checkpoint-65000/transformer/diffusion_pytorch_model.bin.index.json
@@ -0,0 +1,1975 @@
+{
+  "metadata": {
+    "total_size": 41584709888
+  },
+  "weight_map": {
+    "condition_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "condition_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.context_embedder.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.context_embedder.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.norm_out.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.norm_out.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.pos_embed.pos_embed": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.pos_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.proj_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.proj_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.text_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.time_text_embed.timestep_embedder.linear_2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.add_v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_added_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.norm_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_add_out.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_k.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_out.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.attn.to_v.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.0.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff.net.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.0.norm1_context.linear.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.1.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.10.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.11.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.12.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.13.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.14.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.15.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.16.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.17.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.18.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.19.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.2.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.20.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.21.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.22.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_k_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_q_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.add_v_proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_added_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.norm_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_add_out.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_k.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_out.0.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_q.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.attn.to_v.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.ff_context.net.2.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.23.norm1_context.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.bias": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1.linear.weight": "diffusion_pytorch_model-00003-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.24.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.25.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.26.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.27.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.28.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.29.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.3.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.30.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.31.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.32.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.33.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.ff_context.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.34.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_k_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_q_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.add_v_proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_added_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.norm_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_add_out.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_k.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_out.0.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_q.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.attn.to_v.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff.net.2.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.bias": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.35.norm1_context.linear.weight": "diffusion_pytorch_model-00004-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_add_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.ff_context.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.36.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_k_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_q_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.add_v_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_added_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.norm_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_k.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_out.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_q.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.attn.to_v.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.0.proj.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.ff.net.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.37.norm1_context.linear.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.4.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.5.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.6.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.7.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.8.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_k_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_q_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.add_v_proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_added_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.norm_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_add_out.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_k.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_out.0.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_q.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.attn.to_v.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.0.proj.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.ff_context.net.2.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.bias": "diffusion_pytorch_model-00002-of-00005.bin",
+    "dit.transformer_blocks.9.norm1_context.linear.weight": "diffusion_pytorch_model-00002-of-00005.bin",
+    "input_embeds_align_mlp.0.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.0.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "input_embeds_align_mlp.2.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "lmm.lm_head.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.embed_tokens.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.0.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.1.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.10.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.11.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.12.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.13.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.14.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.15.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.16.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.17.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.18.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.19.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.2.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.20.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.21.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.22.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.23.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.24.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.25.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.26.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.27.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.3.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.4.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.5.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.6.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.7.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.8.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.input_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.down_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.gate_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.mlp.up_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.post_attention_layernorm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.k_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.o_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.q_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.layers.9.self_attn.v_proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.model.norm.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.0.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.1.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.10.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.11.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.12.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.13.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.14.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.15.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.16.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.17.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.18.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.19.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.2.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.20.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.21.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.22.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.23.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.24.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.25.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.26.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.27.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.28.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.29.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.3.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.30.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.31.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.4.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.5.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.6.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.7.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.8.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.attn.qkv.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.mlp.fc2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm1.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.blocks.9.norm2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.ln_q.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.0.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.bias": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.merger.mlp.2.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "lmm.visual.patch_embed.proj.weight": "diffusion_pytorch_model-00001-of-00005.bin",
+    "norm_lmm_out.weight": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.bias": "diffusion_pytorch_model-00005-of-00005.bin",
+    "pooled_proj.weight": "diffusion_pytorch_model-00005-of-00005.bin"
+  }
+}