Skip to content

Conversation

@frazane
Copy link
Contributor

@frazane frazane commented Dec 9, 2025

This PR reworks the program used to generate the pyproject.toml file for a given checkpoint.

Summary of changes

  • all dependencies information is extracted from the checkpoint and no longer from MLFlow - a requirements.txt is automatically generated from the checkpoint using anemoi-inference inspect --requirements <path-to-checkpoint>
  • uv is used to add those dependencies to the pyproject
  • the rule "create_inference_pyproject" is updated to use a shell directive instead of a script directive

Note: right now, this requires anemoi-inference to be installed in the evalml virtual environment, which unfortunately comes with some heavy dependencies such as torch. Ideally, commands to interact with the checkpoint should be moved to anemoi-utils. Until then, I would keep this PR as a draft.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants