update

logs/text2image-fine-tune/1688268525.9759955/events.out.tfevents.1688268525.instance-2.1597.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb406acc3ddddc664a802c8df4dbf5e90505f5e02d3dfec504163b3b794a24a7
+size 2201

logs/text2image-fine-tune/1688268525.9774733/hparams.yml

@@ -0,0 +1,50 @@
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_epsilon: 1.0e-08
+adam_weight_decay: 0.01
+allow_tf32: false
+cache_dir: null
+caption_column: text
+center_crop: true
+checkpointing_steps: 500
+checkpoints_total_limit: null
+dataloader_num_workers: 0
+dataset_config_name: null
+dataset_name: JAWCF/characters
+enable_xformers_memory_efficient_attention: false
+gradient_accumulation_steps: 4
+gradient_checkpointing: false
+hub_model_id: null
+hub_token: null
+image_column: image
+input_perturbation: 0
+learning_rate: 1.0e-05
+local_rank: -1
+logging_dir: logs
+lr_scheduler: constant
+lr_warmup_steps: 0
+max_grad_norm: 1.0
+max_train_samples: null
+max_train_steps: 5000
+mixed_precision: null
+noise_offset: 0
+non_ema_revision: null
+num_train_epochs: 200
+output_dir: sd-pokemon-model
+prediction_type: null
+pretrained_model_name_or_path: stabilityai/stable-diffusion-2-1
+push_to_hub: false
+random_flip: true
+report_to: tensorboard
+resolution: 768
+resume_from_checkpoint: null
+revision: null
+scale_lr: false
+seed: null
+snr_gamma: null
+tracker_project_name: text2image-fine-tune
+train_batch_size: 1
+train_data_dir: null
+use_8bit_adam: false
+use_ema: true
+validation_epochs: 5

logs/text2image-fine-tune/events.out.tfevents.1688268525.instance-2.1597.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d37ac6478a46508c2f30a1beb3d20089d2f0982253e5fbb900f302381c0d7071
+size 244961

model_index.json

@@ -5,14 +5,14 @@
     "transformers",
     "CLIPImageProcessor"
   ],
-  "requires_safety_checker": true,
+  "requires_safety_checker": false,
   "safety_checker": [
-    "stable_diffusion",
-    "StableDiffusionSafetyChecker"
+    null,
+    null
   ],
   "scheduler": [
     "diffusers",
-    "PNDMScheduler"
+    "DDIMScheduler"
   ],
   "text_encoder": [
     "transformers",

scheduler/scheduler_config.json

@@ -1,14 +1,20 @@
 {
-  "_class_name": "PNDMScheduler",
+  "_class_name": "DDIMScheduler",
   "_diffusers_version": "0.18.0.dev0",
   "beta_end": 0.012,
   "beta_schedule": "scaled_linear",
   "beta_start": 0.00085,
   "clip_sample": false,
+  "clip_sample_range": 1.0,
+  "dynamic_thresholding_ratio": 0.995,
   "num_train_timesteps": 1000,
-  "prediction_type": "epsilon",
+  "prediction_type": "v_prediction",
+  "rescale_betas_zero_snr": false,
+  "sample_max_value": 1.0,
   "set_alpha_to_one": false,
   "skip_prk_steps": true,
   "steps_offset": 1,
+  "thresholding": false,
+  "timestep_spacing": "leading",
   "trained_betas": null
 }

text_encoder/config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "CompVis/stable-diffusion-v1-4",
+  "_name_or_path": "stabilityai/stable-diffusion-2-1",
   "architectures": [
     "CLIPTextModel"
   ],
@@ -7,16 +7,16 @@
   "bos_token_id": 0,
   "dropout": 0.0,
   "eos_token_id": 2,
-  "hidden_act": "quick_gelu",
-  "hidden_size": 768,
+  "hidden_act": "gelu",
+  "hidden_size": 1024,
   "initializer_factor": 1.0,
   "initializer_range": 0.02,
-  "intermediate_size": 3072,
+  "intermediate_size": 4096,
   "layer_norm_eps": 1e-05,
   "max_position_embeddings": 77,
   "model_type": "clip_text_model",
-  "num_attention_heads": 12,
-  "num_hidden_layers": 12,
+  "num_attention_heads": 16,
+  "num_hidden_layers": 23,
   "pad_token_id": 1,
   "projection_dim": 512,
   "torch_dtype": "float16",

text_encoder/pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a34f30098988d85dc0fb0fc272a842ebcf552e2ebc6ce4adbcf3695d08e8a90
-size 246188833
+oid sha256:f2a06cf32cf585d03b55fef302142a5321b761ec440113925f64f4ceaffc7730
+size 680904225

tokenizer/special_tokens_map.json

@@ -13,7 +13,7 @@
     "rstrip": false,
     "single_word": false
   },
-  "pad_token": "<|endoftext|>",
+  "pad_token": "!",
   "unk_token": {
     "content": "<|endoftext|>",
     "lstrip": false,

unet/config.json

@@ -1,11 +1,16 @@
 {
   "_class_name": "UNet2DConditionModel",
   "_diffusers_version": "0.18.0.dev0",
-  "_name_or_path": "CompVis/stable-diffusion-v1-4",
+  "_name_or_path": "stabilityai/stable-diffusion-2-1",
   "act_fn": "silu",
   "addition_embed_type": null,
   "addition_embed_type_num_heads": 64,
-  "attention_head_dim": 8,
+  "attention_head_dim": [
+    5,
+    10,
+    20,
+    20
+  ],
   "block_out_channels": [
     320,
     640,
@@ -17,7 +22,7 @@
   "class_embeddings_concat": false,
   "conv_in_kernel": 3,
   "conv_out_kernel": 3,
-  "cross_attention_dim": 768,
+  "cross_attention_dim": 1024,
   "cross_attention_norm": null,
   "down_block_types": [
     "CrossAttnDownBlock2D",
@@ -45,7 +50,7 @@
   "resnet_out_scale_factor": 1.0,
   "resnet_skip_time_act": false,
   "resnet_time_scale_shift": "default",
-  "sample_size": 64,
+  "sample_size": 96,
   "time_cond_proj_dim": null,
   "time_embedding_act_fn": null,
   "time_embedding_dim": null,
@@ -57,6 +62,6 @@
     "CrossAttnUpBlock2D",
     "CrossAttnUpBlock2D"
   ],
-  "upcast_attention": false,
-  "use_linear_projection": false
+  "upcast_attention": true,
+  "use_linear_projection": true
 }

unet/diffusion_pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2d28540e6d1458dc018ba11277f20dc1ac38b328e356453317fe97c151d469c1
-size 3438375973
+oid sha256:712ef4aed1c65cca86898dbf038b9e0f1bf362232a15e7003e1f28fb355b4045
+size 3463934693

vae/config.json

@@ -1,7 +1,7 @@
 {
   "_class_name": "AutoencoderKL",
   "_diffusers_version": "0.18.0.dev0",
-  "_name_or_path": "CompVis/stable-diffusion-v1-4",
+  "_name_or_path": "stabilityai/stable-diffusion-2-1",
   "act_fn": "silu",
   "block_out_channels": [
     128,
@@ -20,7 +20,7 @@
   "layers_per_block": 2,
   "norm_num_groups": 32,
   "out_channels": 3,
-  "sample_size": 512,
+  "sample_size": 768,
   "scaling_factor": 0.18215,
   "up_block_types": [
     "UpDecoderBlock2D",

vae/diffusion_pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa8c1b74b3e2781e4347b9b350203597674d8860a4338b46431de760c3a5dd22
+oid sha256:7cfdd672df17db3283633acb3721afc7735927293c2d3bd2bf64939a6dcd950e
 size 167407857