SFT distillation: bug fixes, VLM support, and pretokenization optimization#2053

Merged

samsja merged 26 commits intomainfrom

eli/sft-distillation

Mar 22, 2026

Commits on Mar 3, 2026

SFT distillation with teacher endpoint
willccbb
committed

Commits on Mar 4, 2026

Commits on Mar 5, 2026

chore: apply ruff formatting
willccbb
committed
docs: add changelog entries for RolloutModelConfig, SFTLossConfig, and CustomLossConfig

cursoragent
and
willccbb
committed

Commits on Mar 6, 2026

refine external SFT rollout path and cleanup tests
willccbb
committed

Commits on Mar 7, 2026

refactor shared SFT chat-template tokenization logic
willccbb
committed

Commits on Mar 19, 2026

Commits on Mar 20, 2026