Question on the compilation with different version of cuda or pytorch
PeiqinZhuang opened this issue · comments
Thanks for this work. I just wonder if the Cuda source code could be compiled with different versions of PyTorch or Cuda driver which you recommend? Right now, I have some problems with older pytorch and cuda driver for the compilation.
Hi, thank you for your interest.
It should be compatible with recent PyTorch and CUDA versions. We've only tried torch 1.8 and above, and strongly recommend the latest, torch 1.11.
Could you share the error you're facing and the torch/cuda versions you have?
Hi, thank you for your interest. It should be compatible with recent PyTorch and CUDA versions. We've only tried torch 1.8 and above, and strongly recommend the latest, torch 1.11. Could you share the error you're facing and the torch/cuda versions you have?
Hi, Thanks for the reply. I attach the screenshot with the error. Actually, I use torch 1.8 and Cuda driver 10.1
Hi Peiqin, what GPU are you using? I don't think we have tested that cuda driver but are looking into it.
This is really strange. Seems to be more of a torch error than anything else. Would you mind also sharing the entire error and the command you ran, and as many details as possible on the environment, so we try and see if we can reproduce it on our end?
Hi Peiqin, what GPU are you using? I don't think we have tested that cuda driver but are looking into it.
Hi, I just compiled it on V100.
Hi,
Could you share the command you ran and your environment details so we can look into the issue?
Hi, Could you share the command you ran and your environment details so we can look into the issue?
Hi, I just solved this problem by creating a new python environment. It seems that there exist some problems with my previous environment.
Thank you for your response.
If you don't have any other questions, feel free to close this issue.
Otherwise feel free to ask.
Closing this due to inactivity. If you still have questions feel free to open it back up.