untimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:4 and cuda:0!

[simba0626]

Hi, authors, It is nice work. I meet one problem: forward of the CLIPVisionEmbeddings class in modeling_clip.py

Could you help me to solve it?

Thanks

[KjAeRsTuIsK]

Hi @simba0626 , do you face this error while evaluating or training? If during evaluation, first try running it on one GPU device. And for training, make sure you use zero2.json for deepspeed. Let me know if you still face this issue.

[simba0626]

Hi, author, thank for your repsonse. First I try running it on one GPU device for evaluation using geochat_demo.py.

However, I meet new problem, as follows：

in geochat_arch.py line number 122: if (cur_input_ids == IMAGE_TOKEN_INDEX).sum() == 0:

Thank you