gowitheflow commited on
Commit
108f6f1
·
verified ·
1 Parent(s): f278aee

fix README example code bugs

Browse files
Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -248,13 +248,13 @@ with torch.no_grad():
248
  }
249
  ] for audio in batch_audios]
250
 
251
- text = self.processor.apply_chat_template(
252
  messages, tokenize=False, add_generation_prompt=True
253
  )
254
  audio_inputs, image_inputs, video_inputs = process_mm_info(
255
  messages, use_audio_in_video=False
256
  )
257
- inputs = self.processor(
258
  text=text,
259
  audio=audio_inputs,
260
  images=image_inputs,
@@ -263,7 +263,7 @@ with torch.no_grad():
263
  padding=True
264
  )
265
  inputs = inputs.to("cuda")
266
- audio_outputs = self.model(
267
  **inputs, output_hidden_states=True, return_dict=True
268
  ).hidden_states[-1][:, -1, :]
269
  all_audio_embeddings.append(audio_outputs.to(torch.float16).cpu())
@@ -324,13 +324,13 @@ with torch.no_grad():
324
  }
325
  ] for video in batch_videos]
326
 
327
- text = self.processor.apply_chat_template(
328
  messages, tokenize=False, add_generation_prompt=True
329
  )
330
  audio_inputs, image_inputs, video_inputs = process_mm_info(
331
  messages, use_audio_in_video=False
332
  )
333
- inputs = self.processor(
334
  text=text,
335
  audio=audio_inputs,
336
  images=image_inputs,
@@ -339,7 +339,7 @@ with torch.no_grad():
339
  padding=True
340
  )
341
  inputs = inputs.to("cuda")
342
- video_outputs = self.model(
343
  **inputs, output_hidden_states=True, return_dict=True
344
  ).hidden_states[-1][:, -1, :]
345
  all_video_embeddings.append(video_outputs.to(torch.float16).cpu())
 
248
  }
249
  ] for audio in batch_audios]
250
 
251
+ text = processor.apply_chat_template(
252
  messages, tokenize=False, add_generation_prompt=True
253
  )
254
  audio_inputs, image_inputs, video_inputs = process_mm_info(
255
  messages, use_audio_in_video=False
256
  )
257
+ inputs = processor(
258
  text=text,
259
  audio=audio_inputs,
260
  images=image_inputs,
 
263
  padding=True
264
  )
265
  inputs = inputs.to("cuda")
266
+ audio_outputs = model(
267
  **inputs, output_hidden_states=True, return_dict=True
268
  ).hidden_states[-1][:, -1, :]
269
  all_audio_embeddings.append(audio_outputs.to(torch.float16).cpu())
 
324
  }
325
  ] for video in batch_videos]
326
 
327
+ text = processor.apply_chat_template(
328
  messages, tokenize=False, add_generation_prompt=True
329
  )
330
  audio_inputs, image_inputs, video_inputs = process_mm_info(
331
  messages, use_audio_in_video=False
332
  )
333
+ inputs = processor(
334
  text=text,
335
  audio=audio_inputs,
336
  images=image_inputs,
 
339
  padding=True
340
  )
341
  inputs = inputs.to("cuda")
342
+ video_outputs = model(
343
  **inputs, output_hidden_states=True, return_dict=True
344
  ).hidden_states[-1][:, -1, :]
345
  all_video_embeddings.append(video_outputs.to(torch.float16).cpu())