CarperAI/trlx
A distributed training framework for fine-tuning large language models using Reinforcement Learning via Human Feedback (RLHF).
Not currently ranked — collecting fresh signals.
star history
trlX is a framework designed from the ground up for fine-tuning large language models with reinforcement learning. It supports training via either a provided reward function or reward-labeled datasets. The framework supports distributed training across multiple devices and was published at EMNLP 2023.
Frequently asked
- What is CarperAI/trlx?
- A distributed training framework for fine-tuning large language models using Reinforcement Learning via Human Feedback (RLHF).
- Is trlx open source?
- Yes — CarperAI/trlx is open source, released under the MIT license.
- What language is trlx written in?
- CarperAI/trlx is primarily written in Python.
- How popular is trlx?
- CarperAI/trlx has 4.8k stars on GitHub.
- Where can I find trlx?
- CarperAI/trlx is on GitHub at https://github.com/CarperAI/trlx.