Failed to build from source (pytorch 1.3.1 + CUDA 11.6) #2321
-
|
Hello, Thanks for your attention to my issue. The corresponding environment is shown as follows: Thanks for your help. Hope you have a good day. Btw, I can build ColossalAI successfully under pytorch 1.12.0. I appreciate if you can release the official release version for ColossalAI 0.2.0 + pytorch 1.3.1 |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Hi all, The root cause of this building issue is because of the wrong environment variable. However, in the anaconda virtual environment with Thanks for the great work! |
Beta Was this translation helpful? Give feedback.
Hi all,
Thanks for your attention!
I build from the source successfully!
The root cause of this building issue is because of the wrong environment variable.
In ColossalAI setup pipeline, the script will parse the path of the cuda from
torch.utils.cpp_extension.CUDA_HOME.And the
torch.utils.cpp_extension.CUDA_HOMEget the path based on the commandwhich nvccif there is no proper environment variable. (ref. https://github.com/pytorch/pytorch/blob/master/torch/utils/cpp_extension.py#L90)However, in the anaconda virtual environment with
pytorch-cuda=11.6, there is a duplicatednvccexecute file in thebin/, which makes the commandwhichgrabs the wrong path before grabbing the correct cuda…