search

TencentARC / PhotoMaker

PhotoMaker [CVPR 2024]
https://photo-maker.github.io/
Other
9.58k stars 768 forks source link

I have an AMD graphics card, hope someone could help me. #93

Open Reshex opened 10 months ago

Reshex commented 10 months ago

Whenever I am running gui.bat i am getting this error: Pipelines loaded with dtype=torch.float16 cannot run with cpu device. It is not recommended to move them to cpu as running them will fail. Please make sure to use an accelerator to run the pipeline in inference, due to the lack of support forfloat16 operations on this device in PyTorch. Please, remove the torch_dtype=torch.float16 argument, or use another device for inference.|

Beacuse i am running an AMD gpu the "cuda" option inside the torch_dtype function does not work for me. but sadly i don't know what to change it into.

Moreover when i am going inside my local url and trying to generate an image a different error message pops in terminal: return torch.layer_norm(input, normalized_shape, weight, bias, eps, torch.backends.cudnn.enabled) RuntimeError: "LayerNormKernelImpl" not implemented for 'Half'

hope somebody could help me please <3

Paper99 commented 10 months ago

Hope too..

Reshex commented 10 months ago

Hope too..

Does it mean that there is no option for AMD GPU to run this?

bigcat88 commented 9 months ago

Does it mean that there is no option for AMD GPU to run this?

AMD 7900 XT/XTX runs code from this repo without any source modifications and they are detected as cuda compatible in PyTorch.

Reshex commented 9 months ago

Does it mean that there is no option for AMD GPU to run this?

AMD 7900 XT/XTX runs code from this repo without any source modifications and they are detected as cuda compatible in PyTorch.

I have 6900 XT, Does it mean i have to use modifications? and if so what are those in order for it to work?

bigcat88 commented 9 months ago

Does PyTorch detects your card as cuda? If yes, then it should be compatible.

Something like:

>>> import torch

print(torch.cuda.is_available())
Reshex commented 9 months ago

Does PyTorch detects your card as cuda? If yes, then it should be compatible.

Something like:

>>> import torch

print(torch.cuda.is_available())

I don't think that pytorch detects my card as cuda. That is the main problem

vectris-dev commented 1 week ago

Is it possible to run this on an AMD mi300x? I want to run it on Runpod and specifically need to use AMD hardware.