ai-agent

Trinity-RFT: Fine-Tune LLMs

Name: Trinity-RFT: Fine-Tune LLMs
Author: agentscope-ai

Get flexible reinforcement fine-tuning for large language models with Trinity-RFT, a framework for founders working with AI agents.

662 stars72 forksPythonQuality 8/10Updated 6/30/2026100% free · open source

What it does

Trinity-RFT allows founders to fine-tune large language models using reinforcement learning, enabling more flexible and scalable AI agent development

Install / run

pip install -r requirements.txt

When to use it

•When you need to adapt a pre-trained language model to a specific task or domain
•When you want to improve the performance of a language model on a particular dataset or metric
•When you need to integrate reinforcement learning with other AI techniques, such as supervised or unsupervised learning

Quick start

1Clone the Trinity-RFT repository from GitHub using `git clone https://github.com/agentscope-ai/Trinity-RFT.git`
2Navigate to the repository directory using `cd Trinity-RFT`
3Configure the `config.json` file to specify the language model, dataset, and reinforcement learning parameters
4Run the `train.py` script to start the reinforcement fine-tuning process using `python train.py --config config.json`
5Monitor the training process and evaluate the performance of the fine-tuned model using the `evaluate.py` script

Ready-to-paste prompt

python train.py --config config.json --model-name 't5-base' --dataset 'wikitext-103' --reward 'bleu'

Heads up: Ensure you have the required GPU resources and a compatible CUDA version, as specified in the Trinity-RFT README, to avoid compatibility issues during training

Saves to your device