Hands-on: Converting the PyTorch single GPU AI training job to use all GPUs in a single node via DDP

Exercises on the course GitHub.

Q&A

1. Does cotainr set SINGULARITY_CACHEDIR and _TMPDIR internally?

   • Is your question essentially where the container is build? If yes, then cotainr creates a singularity sandbox in the /tmp folder. It uses the tempfile module built into python to determine the folder.