TRL Post-Training
Quick reference for post-training LLMs with TRL: SFT, DPO, GRPO, KTO and reward models, with the config fields that matter.
Install
- 1
Get Haarness
Free and open source. Skip this step if Haarness is already on your computer.
Free and open source. Skip this step if Haarness is already on your computer.
Free and open source. Skip this step if Haarness is already on your computer.
Haarness installs on a computer. Open this page there to continue, or use it in the cloud below.
Haarness installs on a computer. Open this page there to continue, or use it in the cloud below.
Haarness installs on a computer. Open this page there to continue, or use it in the cloud below.
Haarness installs on a computer. Open this page there to continue, or use it in the cloud below.
- 2
Add TRL Post-Training
Run this in a terminal. Add --global to install for every project, or --local for this project only.
haarness marketplace install skill haarness/trl-training
Notes
From huggingface/skills (skills/trl-training), by huggingface. Licence: Apache-2.0. Haarness keeps it in sync with the repository; installs update themselves.
Details
- Publisher
- Haarness
- Version
- v1
- License
- Apache-2.0
- Visibility
- public
- Updated
- Source
- github.com/huggingface/skillsskills/trl-training · mainKept in sync · 25910b3
Versions
- v1