fix: accelerate not found (modelscope#427)

accelerate may not bye available in some envs in sd-webui,   and in most cases training a lora model only need on gpu, so remove accelerate

Co-authored-by: Xingjun.Wang <[email protected]>

app.py

@@ -120,7 +120,7 @@ def train_lora_fn(base_model_path=None, revision=None, sub_path=None, output_img
             ]
         else:
             command = [
-                'accelerate', 'launch', f'{project_dir}/facechain/train_text_to_image_lora_sdxl.py' if base_model_path is SDXL_BASE_MODEL_ID else f'{project_dir}/facechain/train_text_to_image_lora.py',
+                'python', f'{project_dir}/facechain/train_text_to_image_lora_sdxl.py' if base_model_path is SDXL_BASE_MODEL_ID else f'{project_dir}/facechain/train_text_to_image_lora.py',
                 f'--pretrained_model_name_or_path={base_model_path}',
                 f'--revision={revision}',
                 f'--sub_path={sub_path}',