Upload HF_GPTForCausalLM

config.json

@@ -0,0 +1,29 @@
+{
+  "architectures": [
+    "HF_GPTForCausalLM"
+  ],
+  "bos_token_id": 1,
+  "conv_size": 4,
+  "eos_token_id": 2,
+  "expand_k": 1.0,
+  "expand_v": 1.0,
+  "hidden_size": null,
+  "mode": "fused_chunk",
+  "model_type": "hf_gpt",
+  "name": "Mamba2_1.4B",
+  "norm_eps": 1e-05,
+  "norm_first": false,
+  "num_heads": 4,
+  "pad_token_id": 0,
+  "qk_activation": "silu",
+  "qk_norm": "l2",
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.43.2",
+  "use_beta": true,
+  "use_cache": true,
+  "use_gate": false,
+  "use_output_norm": true,
+  "use_short_conv": true,
+  "vocab_size": 32000
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 0,
+  "transformers_version": "4.43.2"
+}

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e10cb37df9ea4b4ccfd91bd8a32dcc019cf6b9f8e9f5835835923bc28919682e
+size 4988554272

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fddcaed871326dbb46a8af2d88a50b553d4966954e2c56f0e91087e54f3b282
+size 1009028920

model.safetensors.index.json

@@ -0,0 +1,244 @@
+{
+  "metadata": {
+    "total_size": 5997557952
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00002-of-00002.safetensors",
+    "transformer.embeddings.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.mlp.swiglu.w2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.15.mlp.swiglu.w3.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.15.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.16.attn.A_log": "model-00002-of-00002.safetensors",
+    "transformer.h.16.attn.D": "model-00002-of-00002.safetensors",
+    "transformer.h.16.attn.conv1d.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.16.attn.conv1d.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.attn.dt_bias": "model-00002-of-00002.safetensors",
+    "transformer.h.16.attn.in_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.attn.norm.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.attn.out_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.mlp.swiglu.w1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.mlp.swiglu.w2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.mlp.swiglu.w3.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.norm_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.16.norm_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.A_log": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.D": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.conv1d.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.conv1d.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.dt_bias": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.in_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.norm.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.attn.out_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.mlp.swiglu.w1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.mlp.swiglu.w2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.mlp.swiglu.w3.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.norm_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.17.norm_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.2.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.A_log": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.D": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.conv1d.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.conv1d.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.dt_bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.in_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.norm.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.out_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.mlp.swiglu.w1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.mlp.swiglu.w2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.mlp.swiglu.w3.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.norm_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.norm_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.ln_f.weight": "model-00002-of-00002.safetensors"
+  }
+}