Hello, based on the previous posts, I have successfully resolved the issue of sharing a conda environment among multiple users in TLJH. I am very grateful for your help. However, a new problem has arisen. While using the virtual environment under the initially configured admin user, there are no errors, but other users encounter a CUDA-related error. Specifically, the error is “Torch not compiled with CUDA enabled.”
The CUDA installation path is /usr/local/cuda-12.1. What configuration do I need to resolve this issue?
Additionally, I have another question: To provide the virtual environment for multiple users, I installed nb_conda_kernels from conda-forge. However, locally created virtual environments do not show up correctly in the Jupyter Notebook unless they are added to the IPython kernel. Is there a way to keep the shared virtual environment displayed while allowing multiple users to see their own locally created environments in the Jupyter Notebook list?
