Skip to content

Conversation

@finbarrtimbers
Copy link
Collaborator

@finbarrtimbers finbarrtimbers commented Nov 2, 2025

Initial implementation done by Codex.

Previously, our image was ~21.5GiB. Now, it's 17.1GiB.

Summary

  • introduce a dedicated CUDA devel builder stage to resolve Python dependencies with uv
  • copy the prepared virtual environment into a slimmer CUDA runtime stage that installs only runtime packages

Runs:

  1. Single GPU: Beaker
  2. Tool use: Beaker
  3. DPO: Beaker
  4. Multi-node GRPO: Beaker

@finbarrtimbers
Copy link
Collaborator Author

Annoyingly DeepSpeed requires NVCC! Closing this until/if we remove DeepSpeed.

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

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants