# TRL

Provides trainers for SFT, reward modelling, PPO, DPO and related alignment methods.

## What it is used for

Running preference-tuning and RLHF-style pipelines.