TRL (Transformers Reinforcement Learning)
Version 1.14.1
Maintained by Hugging Face14
TRL is Hugging Face's open-source library for post-training transformer language models. It provides trainer classes for methods including supervised fine-tuning (SFT), Group Relative Policy Optimization (GRPO), Direct Preference Optimization (DPO), KTO, and reward modeling, with further methods marked experimental, and it is built on the Transformers library.42
- Repository: Repository (external site: github.com)
- Documentation: Documentation (external site: huggingface.co)
- License: LICENSE (external site: raw.githubusercontent.com)
- Release notes: Release v1.14.1 (external site: github.com)
Availability and license
Public materials checklist
| Item | Status | Notes and evidence |
|---|---|---|
| Source codeIs the source code publicly readable? | Public | 1 |
| DocumentationIs user documentation published? | Public | The documentation covers installation, quickstart, conceptual and how-to guides, integrations (for example DeepSpeed, Liger Kernel, PEFT), and API references.4 |
| InstallationAre installation instructions or packages publicly available? | Public | Documented installation from PyPI with pip or uv, or from source.5 |
| Supported platformsAre supported operating systems or hardware documented? | Unknown | Not assessed. |
| Release statusAre versioned releases published? | Public | Versioned releases are published on GitHub; v1.14.1 was published on 2026-09-29.6 |
What it is useful for
Fine-tuning and aligning language models on custom datasets, from a single GPU to multi-node setups, including parameter-efficient training through PEFT (LoRA/QLoRA) and a command-line interface for training without writing code.2
Run and use notes
- The README states that each TRL trainer wraps the Transformers trainer and supports distributed training methods such as DDP, DeepSpeed ZeRO, and FSDP, using Accelerate to scale from one GPU to multi-node clusters.2
Organization context
U.S. eligibility
Eligible · basis: U.S.-governed project
TRL is developed in Hugging Face's GitHub organization and documented on Hugging Face's documentation site. Hugging Face's terms of service identify Hugging Face, Inc., a Delaware corporation, as the provider of its services, under New York law and courts, and its privacy policy states that the company and its servers are located in the United States (the French entity is named as its EU main establishment). See the hugging-face organization record for the full dual-country assessment.1478
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.