trl

github:evalstate/skills-dev · first committed 2025-10-29 · source · badge

This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, co

Skill Health: not yet scored

Listed, not yet scored. This skill has no public model test yet. It is in the scoring queue; submitted skills go first, then the most-installed.

Version history

11 commits to this file on GitHub (newest first).

2025-10-31 · commit 1722f26 · env vars was deprecated
2025-10-31 · commit 2284312 · trackio project env var
2025-10-31 · commit 89207de · improve trackio integration instructions
2025-10-30 · commit 7f2c699 · trackio improvements
2025-10-29 · commit e621887 · skip eval for demo instructions
2025-10-29 · commit 7093da7 · update preferences prefer uv
2025-10-29 · commit eb23fb4 · remove kto/ppo
2025-10-29 · commit 59eece6 · update dataset validation instructions, remove unused train styles
2025-10-29 · commit a8a2531 · add dataset inspector
2025-10-29 · commit d55cac6 · aksel skill updates
2025-10-29 · commit 6527162 · trl v1 (missing hf merges)

Archived snapshots