Skip to content

v0.3.0

Latest

Choose a tag to compare

@VihangaFTW VihangaFTW released this 13 Mar 11:30
· 2 commits to main since this release

2026-03-13

Changed

  • Training pipeline substantially refactored: all text preprocessing (including regex chunking and pretokenization) is now performed by a parallelized Rust-side workflow. This reduces training time by over 80% and cuts memory usage by approximately 50%.