search

m-tassano / fastdvdnet

FastDVDnet: A Very Fast Deep Video Denoising algorithm
MIT License
572 stars 123 forks source link

nvidia-sali #47

Closed Rose0506 closed 2 years ago

Rose0506 commented 2 years ago

Hi,Matias~I have a question about the using of nvidia to ask you for help.Although I have install a corresonding version of nvidia-dali,when I trained the model, AttributeError: module 'nvidia.dali.ops' has no attribute 'CropCastPermute' always occurs.I have no idea about how to solve it after I tried to install all kinds of vesions of nvidia-sali.It appears to be torch model without permute function.

m-tassano commented 2 years ago

Hi, I think you're using an old version of FastDVDnet. There was an update about a year ago which replaced DALI's CropCastPermute by CropMirrorNormalize, as the former is deprecated. If you use this repo's latest version together with DALI's version in the requirements.yml file, you won't get this error

Rose0506 commented 2 years ago

Thanks very much.The project i use is about video snap compression where your old version of FastDVDNet is used for the denoising of the reconstrcution video by the author.Unfortunately,after solving this problem,another one arised,when i trained your fastdvdnet by the train_fastdvdnet.py,error diplayed as "RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument mat2 in method wrapper_mm)".I think it is unbelievable,for the reason that all the code has already debugged the code on the GPU.I am looking forward to your help.Thank you very much!

m-tassano commented 2 years ago

I have never encountered this problem. You can try reading this pytorch thread