Upload 3 files

args.json

@@ -0,0 +1,52 @@
+{
+  "guidance_scale_img": 4.5,
+  "guidance_scale_text": 7.5,
+  "config_p": "./configs/VITONHD_agnidmhand_agnv9hand_randdilate_gpt4ov6v12_jointcond_nopose_noip_intermcloth.yaml",
+  "use_sd15": false,
+  "resume_path": null,
+  "pretrained_unet_path": null,
+  "pretrained_unet_before_path": null,
+  "gradient_accumulation_steps": 1,
+  "prediction_type": "epsilon",
+  "timestep_spacing": "leading",
+  "noise_offset": 0.05,
+  "num_inference_steps": 30,
+  "finetuning": false,
+  "save_step": 15000,
+  "validation_step": 30000,
+  "validation_step_lst": [
+    2,
+    30000,
+    60000
+  ],
+  "skip_paired": false,
+  "skip_unpaired": false,
+  "n_samples_train": 2,
+  "n_samples_valid": 999999,
+  "save_root_dir": "./logs",
+  "save_name": "VITONHD_agnidmhand_agnv9hand_randdilate_gpt4ov6v12_jointcond_nopose_noip_intermcloth_bs32",
+  "strength": 1.0,
+  "img_h": 1024,
+  "img_w": 768,
+  "use_8bit_adam": false,
+  "no_zero_snr": false,
+  "init_model_path": "./pretrained_models/stable-diffusion-xl-1.0-inpainting-0.1",
+  "init_vae_path": "./pretrained_models/sdxl-vae-fp16-fix",
+  "init_ip_adapter_path": "./pretrained_models/IP-Adapter/sdxl_models/ip-adapter_sdxl.bin",
+  "init_image_encoder_path": "./pretrained_models/IP-Adapter/sdxl_models",
+  "init_cloth_encoder_path": "./pretrained_models/stable-diffusion-xl-base-1.0",
+  "ip_adapter_num_tokens": 4,
+  "logging_dir": "wandb",
+  "learning_rate": 3.2e-05,
+  "lr_scheduler": "constant_with_warmup",
+  "no_lr_scheduler_accelerator": false,
+  "lr_warmup_steps": 500,
+  "lr_num_cycles": 1,
+  "lr_power": 1.0,
+  "weight_decay": 0.01,
+  "guidance_scale": 2.0,
+  "num_workers": 2,
+  "mixed_precision": "fp16",
+  "report_to": "wandb",
+  "total_batch_size": 32
+}

config.yaml

@@ -0,0 +1,51 @@
+no_pose: true
+use_jointcond: true
+no_ipadapter: true
+use_interm_cloth_mask: true
+repaint_start_ratio: 0.5
+dataset:
+  dataset_name: VITONHDDataset
+  data_root_dir: ./DATA/zalando-hd-resized
+  img_spatial_transform_lst:
+  - randomresizedcrop
+  - randomaffine
+  cloth_spatial_transform_lst:
+  - randomresizedcrop
+  - randomaffine
+  img_cloth_spatial_transform_lst:
+  - hflip
+  color_transform_lst:
+  - colorjitter
+  i_drop_rate: 0.05
+  pose_type: densepose
+  train_folder_name: train_agnidm_sapiens_hand
+  test_folder_name: test_agnidm_sapiens_hand
+  prompt_version: v12
+  text_file_postfix: gpt4o_v6.json
+  train_folder_name_for_interm_cloth_mask: train_agnv9_sapiens_hand
+  test_folder_name_for_interm_cloth_mask: test_agnv9_sapiens_hand
+  use_rand_dilate: true
+  train_min_dilate_folder_name: train_agnidm_sapiens_hand
+  train_max_dilate_folder_name: train_agnv9_sapiens_hand
+  rand_dilate_miniter: 0
+  rand_dilate_maxiter: 200
+extra_train_dataset:
+  dataset1:
+    dataset_name: VITONHDDataset
+    data_root_dir: ./DATA/zalando-hd-resized
+    img_spatial_transform_lst:
+    - randomresizedcrop
+    - randomaffine
+    cloth_spatial_transform_lst:
+    - randomresizedcrop
+    - randomaffine
+    img_cloth_spatial_transform_lst:
+    - hflip
+    color_transform_lst:
+    - colorjitter
+    i_drop_rate: 0.05
+    pose_type: densepose
+    train_folder_name: train_agnv9_sapiens_hand
+    test_folder_name: test_agnv9_sapiens_hand
+    prompt_version: v12
+    text_file_postfix: gpt4o_v6.json

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c01a173d0fc1892fe75730c3cf15bf9d81b7e80b41c4cf1fa49a1cb20bfb6dd
+size 10270149870