Skip to content

Update docker image with latest packages

Minh Tuan Pham requested to merge tp_update_docker into dev
  • Update Dockerfile to decrease the number of RUN commands, hence image size
  • Update requirements.txt to PyTorch 2.4.0+cuda12.1, torch_lightning 2.4.0. Add other packages not explicitly present: wandb, scikit-learn.
  • Docker image built from docker maintained pytorch devel base image
  • In .test-gitlab-ci.yaml, get rid of activation of conda environment. All packages are install into base env and there is no need to activate. Disable wandb by export WANDB_DISABLED=true, seems to work better for current wandb version.
  • In tests/stages/edge_classifier/test_gnn_config.yaml, add log_wandb: false to explicitly turn off wandb logging for test

I switched the merging to branch nightly, to have a buffer before merging to dev. We should consider following this practice to make sure that the dev is protected from bad merge requests.

Edited by Minh Tuan Pham

Merge request reports