Open adamltyson opened 1 year ago
conda will be moving to the mamba solver so this means we should be able to just document miniconda and ignore the complexities of mamba et al.
We also discovered that creating a conda environment with the miniconda
module during an interactive job (started via srun -p cpu --pty bash -i
) is very slow. Creating the same environment on the gateway node seems to be faster. If confirmed, we should solve this or at least document it.
There are many issues about documenting the use of conda. I think this means we should actually do it.