M-Courtyard v0.5.1: Training History, Run Comparison, and Reliable Result Persistence

We’ve released M-Courtyard v0.5.1 for macOS Apple Silicon.

This update focuses on experiment traceability for local fine-tuning:

  • Added a dedicated Training History workspace
  • Added side-by-side run comparison
  • Persisted training results reliably from the Rust backend
  • Fixed missing / inconsistent loss curves in history and comparison mode
  • Improved status accuracy for completed and stopped runs

The goal of this release is to make local fine-tuning workflows more reproducible and easier to iterate on without leaving the desktop app.

GitHub Release:

1 Like