Metadata-Version: 2.4
Name: grporlhf
Version: 0.1.0
Summary: Reference implementation of Group Relative Policy Optimization (GRPO)
Home-page: https://github.com/kaushikd24/GRPO-Package
Author: Kaushik D. (Not associated with DeepSeek -- Demo Implementation)
Author-email: "Kaushik D." <dwivedi.kaushik24@gmail.com>
License: Apache-2.0
Requires-Python: >=3.9
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: torch<2.4,>=2.2
Requires-Dist: transformers<4.43,>=4.41.2
Requires-Dist: peft>=0.11.0
Requires-Dist: datasets>=2.20.0
Requires-Dist: pyyaml
Requires-Dist: tqdm
Dynamic: author
Dynamic: home-page
Dynamic: license-file
Dynamic: requires-python

# GRPO RLHF

This package provides a reference implementation of Group Relative Policy Optimization (GRPO) by DeepSeek. It includes components such as a trainer, CLI, and utilities for handling datasets.

## Installation

You can install the package using pip:

```
pip install grporlhf
```

## Usage

To use the CLI, run:

```
grporlhf --help
```

## License

This project is licensed under the Apache-2.0 License. 
