openai/lm-human-preferences
Code for training reward models and fine-tuning language models from human preferences.
Not currently ranked — collecting fresh signals.
star history
This repository implements the RLHF-style training pipeline from OpenAI’s 2019 paper. It trains reward models from human preference labels, then fine-tunes language models (GPT-2) using those reward models via policy gradient methods. The code includes utilities for distributed training with Horovod and was tested on GPU clusters.
Frequently asked
- What is openai/lm-human-preferences?
- Code for training reward models and fine-tuning language models from human preferences.
- Is lm-human-preferences open source?
- Yes — openai/lm-human-preferences is open source, released under the MIT license.
- What language is lm-human-preferences written in?
- openai/lm-human-preferences is primarily written in Python.
- How popular is lm-human-preferences?
- openai/lm-human-preferences has 1.4k stars on GitHub.
- Where can I find lm-human-preferences?
- openai/lm-human-preferences is on GitHub at https://github.com/openai/lm-human-preferences.