microsoft / LLaVA-Med

Large Language-and-Vision Assistant for Biomedicine, built towards multimodal GPT-4 level capabilities.
Other
1.29k stars 148 forks source link

CLIPVisionTower'object has no attribute 'device' #75

Open 2927803072 opened 1 month ago

2927803072 commented 1 month ago

Hello, When I was running the code for your Medical VQA section, I got the following problem: CLIPVisionTower'object has no attribute 'device'. I don't know if it's a problem with my data format, but here are some of the contents in my test.json: {"id": "image_10", "image": "image_10.jpg", "conversations": [{"from": "human", "value": "What it is in the picture?"}, {"from": "gpt", "value": "Dog."}]}。 Do you know what the reason is? I really hope you can answer my confusion, thank you