Install problem

[XuHangkun]

My cuda version is 11.7 (checked with command "nvcc --version"), and I run "conda env create -f environment.yaml" to build the environment and run "conda activate detectorfreesfm" without any problem. But when I run "conda activate detectorfreesfm", the system show the errors below.

I found the cuda version of install torch is 12.1 with command "torch.version.cuda". So is there any advice to solve this problem?

[hxy-123]

It seems that the installed pytorch version used CUDA 12.1, where the required pytorch version is 1.8.0 and CUDA version is 11.7 in our environment. You can check whether the installed pytorch version is right.

[XuHangkun]

It seems that the installed pytorch version used CUDA 12.1, where the required pytorch version is 1.8.0 and CUDA version is 11.7 in our environment. You can check whether the installed pytorch version is right.

It seems the torch version is changed after the pytorch-lightning is installed. The problem is same as this issue "#38"

[hxy-123]

It seems that the installed pytorch version used CUDA 12.1, where the required pytorch version is 1.8.0 and CUDA version is 11.7 in our environment. You can check whether the installed pytorch version is right.

It seems the torch version is changed after the pytorch-lightning is installed. The problem is same as this issue "#38"

I think it is weird that pytorch-lightning with version 1.3.5 tries to upgrade pytorch version to 2.3.0, as they are not compatible as shown in official manual. You can have a look at whether the final installed pytorch-lightning version is 1.3.5, as required by repo.