Update videollama2/__init__.py

videollama2/__init__.py

@@ -24,7 +24,7 @@ def model_init(model_path=None, **kwargs):
         tokenizer.pad_token = tokenizer.unk_token
 
     num_frames = model.config.num_frames if hasattr(model.config, "num_frames") else NUM_FRAMES
-
+    
     processor = {
         'image': partial(process_image, processor=processor, aspect_ratio=None),
         'video': partial(process_video, processor=processor, aspect_ratio=None, num_frames=num_frames),
@@ -46,9 +46,7 @@ def mm_infer(url, instruct, model, tokenizer, modal='video', **kwargs):
     Returns:
         str: response of the model.
     """
-    image_or_video = 'Temp.mp4'
-    gdown.download(url, output, quiet=False)
-    logging.info(f"video downloaded form: {url}")
+    
     # 1. text preprocess (tag process & generate prompt).
     if modal == 'image':
         modal_token = DEFAULT_IMAGE_TOKEN