Home Columbia Ginsburg HPC Doc
Post
Cancel

Columbia Ginsburg HPC Doc

Content

  1. login
  2. use conda to install packages

login and initiate modules

Refer to the official documentation

significant modules that needs to be loaded

  • module load anaconda/3-2023.09
  • module load cuda12.0/toolkit
  • module load gcc/10.2.0
  • module load cudnn8.0-cuda11.1

Check the version of the loaded modules

common issues

CUDA

1
/usr/lib/ld or /compiler_compat/ld: cannot find -lcuda: No such file or directory

This is because the system cannnot find libcuda.so which is normally located under /cm/shared/apps/cuda11.1/toolkit/11.7.1/lib64/stubs/.

It can be fixed by adding this to LIBRARY_PATH:

1
export LIBRARY_PATH=/cm/shared/apps/cuda11.1/toolkit/11.7.1/lib64/stubs/:$LIBRARY_PATH

CONDA ENV

To activate conda environment in job bash file:

1
2
3
4
5
module load anaconda/3-2023.09
source activate /absolute/path/to/conda/env/
export PATH=/absolute/path/to/conda/env/bin:$PATH

python <python-to-execute.py>
This post is licensed under CC BY 4.0 by the author.
Trending Tags