Differences

This shows you the differences between two versions of the page.

--- doku:python [2024/02/22 16:38] – katrin
+++ doku:python [2024/02/29 09:25] – katrin
@@ Line 72: / Line 72: @@
 </code>
-==== More info about Conda ====
+===== More info about Conda =====
 For more information about conda check-out the conda notebook of the python4HPC training material: [[https://gitlab.tuwien.ac.at/vsc-public/training/python4hpc/-/blob/main/D1_02_env_03_conda.ipynb|python4HPC Conda Environments]]
-==== FAQ ====
+===== FAQ =====
   * **'''eval "$(conda shell.bash hook)"''' fails with '''CommandNotFoundError ...'''**
@@ Line 82: / Line 82: @@
 Make sure to use the currently provided '''miniconda3''' module via '''module load miniconda3'''
-  * **I dont get the right python version and packages** / **I see the error '''ModuleNotFoundError: No module named "conda"''' in my slurm job logfile** - **What is wrong>?**
+  * **I dont get the right python version and packages in my slurm batch file environment** - **What is wrong>?**
-The default for sbatch is to use the current user environment and start the job in the context of this environment. Since conda is depending on environment variables being setup correctly you need to make sure to start the job script from a clean environment.
+Make sure that one of the first things in your sbatch script is loading the miniconda3 package '''module load miniconda3''' and that you execute '''eval "$(conda shell.bash hook)"''' before using any other conda commands.
+  * **I don't get GPU/CUDA enabled packages when installing a conda environment** / **I get conda installation errors when i select GPU/CUDA enabled builds for my conda environment***
+In order to install packages from conda-forge that require CUDA on a machine that does not have GPUs you have to set an environment variable before creating the conda environment:
+<code>
+name: my-pytorch-gpu-env
+channels:
+  - pytorch
+  - conda-forge
+  - defaults
+dependencies:
+  - python=3.12
+  - pytorch=2.2.*=*cuda11.8*
+</code>
+<code>
+# for example if you use cuda 11.8
+CONDA_OVERRIDE_CUDA="11.8" conda env create -n my-pytorch-gpu-env --file my-pytorch-gpu-env.yaml
+</code>
-Alternatively you can also use sbatch with the '''--export=NONE''' argument from commandline or add an sbatch directive in your job script: "#SBATCH --export=NONE"
 ====== Deprecated Information (work in progress) ======