aa

app.py

@@ -41,7 +41,7 @@ os.makedirs("pretrained/v2a/mmaudio", exist_ok=True)
 setup_eval_logging()
 pipeline = Pipeline(
     step0_model_dir=repo_local_path, 
-    step1_mode='mmaudio_medium_44k', 
+    step1_mode='mmaudio_small_44k', 
     step2_model_dir=repo_local_path,
     step2_mode='cot',
     step3_mode='bs_roformer',

third_party/VideoLLaMA2/videollama2/model/__init__.py

@@ -52,7 +52,7 @@ VLLMConfigs = {
 
 
 
-def load_pretrained_model(model_path, model_base, model_name, load_8bit=True, load_4bit=False, device_map="auto", device="cuda", use_flash_attn=False, **kwargs):
+def load_pretrained_model(model_path, model_base, model_name, load_8bit=False, load_4bit=False, device_map="auto", device="cuda", use_flash_attn=False, **kwargs):
     print("00000000000000000000000000: ", device, use_flash_attn)
     if 'token' in kwargs:
         token = kwargs['token']