Update run.py

run.py

@@ -35,33 +35,31 @@ def fast_caption(sys_prompt, usr_prompt, temp, top_p, max_tokens, model, key, en
     elif video_hf and video_hf_auth:
         # Process all videos in the dataset
         all_captions = []
-        with tempfile.NamedTemporaryFile(mode='w+t', delete=True) as temp_parquet_file:
-            temp_parquet_file.write(hf_hub_download(
-                repo_id=video_hf,
-                filename='data/' + str(parquet_index).zfill(6) + '.parquet',
-                repo_type="dataset",
-                token=video_hf_auth,
-            ))
-            parquet_path = temp_parquet_file.name
-            print(parquet_path)
-            parquet_file = pq.ParquetFile(parquet_path)
-            
-            for batch in parquet_file.iter_batches(batch_size=1):
-                df = batch.to_pandas()
-                video = df['video'][0]
+        temp_parquet_file = hf_hub_download(
+            repo_id=video_hf,
+            filename='data/' + str(parquet_index).zfill(6) + '.parquet',
+            repo_type="dataset",
+            token=video_hf_auth,
+        )
+        print(temp_parquet_file)
+        parquet_file = pq.ParquetFile(parquet_path)
+        
+        for batch in parquet_file.iter_batches(batch_size=1):
+            df = batch.to_pandas()
+            video = df['video'][0]
 
-                md5 = hashlib.md5(video).hexdigest()
-                with tempfile.NamedTemporaryFile(mode='w+t', delete=True) as temp_video_file:
-                    temp_video_file.write(video)
-                    video_path = temp_video_file.name
-                    print(video_path)
+            md5 = hashlib.md5(video).hexdigest()
+            with tempfile.NamedTemporaryFile(mode='w+t', delete=True) as temp_video_file:
+                temp_video_file.write(video)
+                video_path = temp_video_file.name
+                print(video_path)
 
-                    processor = VideoProcessor(frame_format=frame_format, frame_limit=frame_limit)
-                    frames = processor._decode(video_path)
-                    base64_list = processor.to_base64_list(frames)
-                    api = AzureAPI(key=key, endpoint=endpoint, model=model, temp=temp, top_p=top_p, max_tokens=max_tokens)
-                    caption = api.get_caption(sys_prompt, usr_prompt, base64_list)
-                    all_captions.append(caption)
+                processor = VideoProcessor(frame_format=frame_format, frame_limit=frame_limit)
+                frames = processor._decode(video_path)
+                base64_list = processor.to_base64_list(frames)
+                api = AzureAPI(key=key, endpoint=endpoint, model=model, temp=temp, top_p=top_p, max_tokens=max_tokens)
+                caption = api.get_caption(sys_prompt, usr_prompt, base64_list)
+                all_captions.append(caption)
 
         return "\n\n\n".join(all_captions), f"Processed {len(video_paths)} videos.", None