- Fix for conversion tool issue when the source was an sd1.x diffuser model

- Other minor code and GUI fix

README.md

@@ -14,6 +14,9 @@ You can find the finetune solution spercific [Finetune README](README_finetune.m
 
 ## Change history
 
+* 12/23 (v18.8) update:
+    - Fix for conversion tool issue when the source was an sd1.x diffuser model
+    - Other minor code and GUI fix
 * 12/22 (v18.7) update:
     - Merge dreambooth and finetune is a common GUI
     - General bug fixes and code improvements

README_dreambooth.md

@@ -67,12 +67,6 @@ python .\tools\cudann_1.8_install.py
 
 When a new release comes out you can upgrade your repo with the following command:
 
-```
-.\upgrade.bat
-```
-
-alternatively you can do it manually with
-
 ```powershell
 cd kohya_ss
 git pull
@@ -87,15 +81,8 @@ Once the commands have completed successfully you should be ready to use the new
 There is now support for GUI based training using gradio. You can start the complete kohya training GUI interface by running:
 
 ```powershell
-.\kohya.cmd
-```
-
-and select the Dreambooth tab.
-
-Alternativelly you can use the Dreambooth focus GUI with
-
-```powershell
-.\dreambooth.cmd
+.\venv\Scripts\activate
+.\kohya_gui.cmd
 ```
 
 ## CLI

README_finetune.md

@@ -67,12 +67,6 @@ python .\tools\cudann_1.8_install.py
 
 When a new release comes out you can upgrade your repo with the following command:
 
-```
-.\upgrade.bat
-```
-
-or you can do it manually with
-
 ```powershell
 cd kohya_ss
 git pull
@@ -110,15 +104,8 @@ You can also use the `Captioning` tool found under the `Utilities` tab in the GU
 There is now support for GUI based training using gradio. You can start the complete kohya training GUI interface by running:
 
 ```powershell
-.\kohya.cmd
-```
-
-and select the Finetune tab.
-
-Alternativelly you can use the Finetune focus GUI with
-
-```powershell
-.\finetune.cmd
+.\venv\Scripts\activate
+.\kohya_gui.cmd
 ```
 
 ## CLI

dreambooth.cmd

@@ -1 +0,0 @@
-.\venv\Scripts\python.exe .\dreambooth_gui.py

finetune.cmd

@@ -1 +0,0 @@
-.\venv\Scripts\python.exe .\finetune_gui.py

finetune/make_captions.py

@@ -24,10 +24,13 @@ def main(args):
   random.seed(seed)
     
   if not os.path.exists("blip"):
+    args.train_data_dir = os.path.abspath(args.train_data_dir)        # convert to absolute path
+
     cwd = os.getcwd()
     print('Current Working Directory is: ', cwd)
     os.chdir('finetune')
 
+  print(f"load images from {args.train_data_dir}")
   image_paths = glob.glob(os.path.join(args.train_data_dir, "*.jpg")) + \
       glob.glob(os.path.join(args.train_data_dir, "*.png")) + glob.glob(os.path.join(args.train_data_dir, "*.webp"))
   print(f"found {len(image_paths)} images.")
@@ -105,4 +108,4 @@ if __name__ == '__main__':
   if args.caption_extention is not None:
     args.caption_extension = args.caption_extention
 
-  main(args)
+  main(args)

kohya.cmd

@@ -1 +0,0 @@
-.\venv\Scripts\python.exe .\kohya_gui.py

tools/convert_diffusers20_original_sd.py

@@ -9,7 +9,7 @@ import os
 import torch
 from diffusers import StableDiffusionPipeline
 
-from library import model_util as model_util
+import library.model_util as model_util
 
 
 def convert(args):
@@ -48,7 +48,7 @@ def convert(args):
       v2_model = unet.config.cross_attention_dim == 1024
       print("checking model version: model is " + ('v2' if v2_model else 'v1'))
     else:
-      v2_model = args.v1
+      v2_model = not args.v1
 
   # 変換して保存する
   msg = ("checkpoint" + ("" if save_dtype is None else f" in {save_dtype}")) if is_save_ckpt else "Diffusers"
@@ -90,4 +90,4 @@ if __name__ == '__main__':
                       help="model to save: checkpoint (with extension) or Diffusers model's directory (without extension) / 変換後のモデル、拡張子がある場合はcheckpoint、ない場合はDiffusesモデルとして保存")
 
   args = parser.parse_args()
-  convert(args)
+  convert(args)

train_db.py

@@ -1011,6 +1011,7 @@ def train(args):
       if stop_text_encoder_training:
         print(f"stop text encoder training at step {global_step}")
         text_encoder.train(False)
+        text_encoder.requires_grad_(False)
 
       with accelerator.accumulate(unet):
         with torch.no_grad():
@@ -1225,4 +1226,4 @@ if __name__ == '__main__':
                       help="Number of steps for the warmup in the lr scheduler (default is 0) / 学習率のスケジューラをウォームアップするステップ数（デフォルト0）")
 
   args = parser.parse_args()
-  train(args)
+  train(args)

upgrade.bat

@@ -1,2 +0,0 @@
-git pull
-.\venv\Scripts\python.exe -m pip install -U -r .\requirements.txt