qwen2-vl-2b-instruct版本微调自己的数据集之后，训练正常，推理报错：IndexError: index 1 is out of bounds for dimension 0 with size 1 #599
Comments
|
将设置的prompt改成这句prompt，Describe this image. 就正常了，感觉很奇怪 |
|
I got this similar issue: However, this is due to the wrong batching setting, i.e., the bz of image and text are inconsistent. By fixing the batching setting, the error can be fixed. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
IndexError Traceback (most recent call last)
Cell In[1], line 41
37 text = processor.apply_chat_template(
38 messages, tokenize=False, add_generation_prompt=True
39 )
40 image_inputs, video_inputs = process_vision_info(messages)
---> 41 inputs = processor(
42 text=[text],
43 images=image_inputs,
44 videos=video_inputs,
45 padding=True,
46 return_tensors="pt",
47 )
48 inputs = inputs.to("cuda")
49 # Inference: Generation of the output
File ~/miniconda3/lib/python3.12/site-packages/transformers/models/qwen2_vl/processing_qwen2_vl.py:137, in Qwen2VLProcessor.call(self, images, text, videos, **kwargs)
134 for i in range(len(text)):
135 while "<|image_pad|>" in text[i]:
136 text[i] = text[i].replace(
--> 137 "<|image_pad|>", "<|placeholder|>" * (image_grid_thw[index].prod() // merge_length), 1
138 )
139 index = 1
140 text[i] = text[i].replace("<|placeholder|>", "<|image_pad|>")
IndexError: index 1 is out of bounds for dimension 0 with size 1
The text was updated successfully, but these errors were encountered: