Skip to content

Conversation

@yayashuxue
Copy link
Contributor

Summary

This PR adds macOS compatibility to rLLM by conditionally excluding GPU-specific dependencies that are not compatible with macOS.

Changes

  • Added sys_platform != 'darwin' conditions to flash-attn, deepspeed, and vllm dependencies
  • Added optional [gpu] dependency group for explicit GPU installation on compatible systems
  • Updated README.md with macOS installation instructions

Benefits

  • ✅ Enables rLLM installation on macOS without GPU dependencies
  • ✅ Maintains full functionality for Linux/CUDA users
  • ✅ Provides backward compatibility
  • ✅ Minimal changes, easy to maintain

Testing

  • Tested installation on macOS (darwin) - ✅ Works
  • Verified Linux compatibility is preserved - ✅ No breaking changes
  • GPU features can still be installed explicitly on compatible systems - ✅ Optional group works

Related Issues

Fixes installation issues on macOS systems where GPU dependencies are not available or compatible.

@jeffreysijuntan
Copy link
Contributor

This is great. We recently updated our README so there is a merge conflict. Can you fix the conflict and then I will merge it in?

@yayashuxue yayashuxue force-pushed the macos-compatibility branch from c1d07a0 to ecc7aa3 Compare August 8, 2025 06:57
@yayashuxue
Copy link
Contributor Author

Fixed! Ready to merge.

@jeffreysijuntan jeffreysijuntan merged commit 44aac5f into rllm-org:main Aug 8, 2025
1 check passed
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