Categories: Software, Usage Tags: Alex Cluster, Basic HPC Knowledge, CUDA, GPU usage, Python/Conda, Slurm, TinyGPU Cluster
Index
Why is my pytorch/tensorflow using CPU only?
The most common mistake is to build the conda/python-venv environment on a fontend and not on a cluster node with GPU.
You can start an interactive job on Alex with salloc --gres=gpu:a40:1 --partition=a40 --time=02:00:00
and on TinyGPU with salloc.tinygpu --gres=gpu:1 --time=02:00:00