update

Author: DN6

examples/advanced_diffusion_training/train_dreambooth_lora_sd15_advanced.py

@@ -1290,6 +1290,7 @@ def save_model_hook(models, weights, output_dir):
                         text_encoder_one_lora_layers_to_save = convert_state_dict_to_diffusers(
                             get_peft_model_state_dict(model)
                         )
+                else:
                     raise ValueError(f"unexpected save model: {model.__class__}")
 
                 # make sure to pop weight so that corresponding model is not saved again
@@ -1981,7 +1982,7 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
         lora_state_dict = load_file(f"{args.output_dir}/pytorch_lora_weights.safetensors")
         peft_state_dict = convert_all_state_dict_to_peft(lora_state_dict)
         kohya_state_dict = convert_state_dict_to_kohya(peft_state_dict)
-        save_file(kohya_state_dict, f"{args.output_dir}/{args.output_dir}.safetensors")
+        save_file(kohya_state_dict, f"{args.output_dir}/kohya_lora_weights.safetensors")
 
         save_model_card(
             model_id if not args.push_to_hub else repo_id,