Trinity-RFT is a general-purpose, flexible, scalable and user-friendly framework designed for reinforcement fine-tuning (RFT) of large language models (LLM).

Built with a decoupled design, seamless integration for agent-environment interaction, and systematic data processing pipelines, Trinity-RFT can be easily adapted for diverse application scenarios, and serve as a unified platform for exploring advanced reinforcement learning (RL) paradigms.

Vision of this project

Current RFT approaches, such as RLHF (Reinforcement Learning from Human Feedback) with proxy reward models or training long-CoT reasoning models with rule-based rewards, are limited in their ability to handle dynamic, real-world, and continuous learning.

Trinity-RFT envisions a future where AI agents learn by interacting directly with environments, collecting delayed or complex reward signals, and continuously refining their behavior through RL.

For example, imagine an AI scientist that designs an experiment, executes it, waits for feedback (while working on other tasks concurrently), and iteratively updates itself based on true environmental rewards when the experiment is finally finished.

Trinity-RFT offers a path into this future by providing various useful features.

Key features

• Unified RFT modes & algorithm support. Trinity-RFT unifies and generalizes existing RFT methodologies into a flexible and configurable framework, supporting synchronous/asynchronous, on-policy/off-policy, and online/offline training, as well as hybrid modes that combine them seamlessly into a single learning process.

• Agent-environment interaction as a first-class citizen. Trinity-RFT allows delayed rewards in multi-step/time-lagged feedback loops, handles long-tailed latencies and environment/agent failures gracefully, and supports distributed deployment where explorers and trainers can operate across separate devices and scale up independently.

• Data processing pipelines optimized for RFT with diverse/messy data. These include converting raw datasets to task sets for RL, cleaning/filtering/prioritizing experiences stored in the replay buffer, synthesizing data for tasks and experiences, offering user interfaces for human in the loop, etc.

The design of Trinity-RFT

The overall design of Trinity-RFT exhibits a trinity:

• RFT-core;
• agent-environment interaction;
• data processing pipelines;

and the design of RFT-core also exhibits a trinity:

• explorer;
• trainer;
• buffer.

The explorer, powered by the rollout model, interacts with the environment and generates rollout trajectories to be stored in the experience buffer.

The trainer, powered by the policy model, samples batches of experiences from the buffer and updates the policy model via RL algorithms.

These two can be completely decoupled and act asynchronously on separate machines, except that they share the same experience buffer, and their model weights are synchronized once in a while. Such a decoupled design is crucial for making the aforementioned features of Trinity-RFT possible.

Meanwhile, Trinity-RFT has done a lot of work to ensure high efficiency and robustness in every component of the framework, e.g., utilizing NCCL (when feasible) for model weight synchronization, sequence concatenation with proper masking for multi-turn conversations and ReAct-style workflows, pipeline parallelism for the synchronous RFT mode, asynchronous and concurrent LLM inference for rollout, fault tolerance for agent/environment failures, among many others.

Getting started

Note

This project is currently under active development. Comments and suggestions are welcome!

Step 1: preparations

Installation from source (recommended):

# Pull the source code from GitHub
git clone https://github.com/modelscope/Trinity-RFT
cd Trinity-RFT

# Create a new environment using Conda or venv
# Option 1: Conda
conda create -n trinity python=3.10
conda activate trinity

# Option 2: venv
python3.10 -m venv .venv
source .venv/bin/activate

# Install the package in editable mode
# for bash
pip install -e .[dev]
# for zsh
pip install -e .\[dev\]

# Install flash-attn after all dependencies are installed
# Note: flash-attn will take a long time to compile, please be patient.
pip install flash-attn -v
# Try the following command if you encounter errors during installation
# pip install flash-attn -v --no-build-isolation

Installation using pip:

pip install trinity-rft==0.1.0

Installation from docker: we have provided a dockerfile for Trinity-RFT (trinity)

git clone https://github.com/modelscope/Trinity-RFT
cd Trinity-RFT

# build the docker image
# Note: you can edit the dockerfile to customize the environment
# e.g., use pip mirrors or set api key
docker build -f scripts/docker/Dockerfile -t trinity-rft:latest .

# run the docker image
docker run -it --gpus all --shm-size="64g" --rm -v $PWD:/workspace -v <root_path_of_data_and_checkpoints>:/data trinity-rft:latest

Trinity-RFT requires Python version >= 3.10, CUDA version >= 12.4, and at least 2 GPUs.

Step 2: prepare dataset and model

Trinity-RFT supports most datasets and models from Huggingface and ModelScope.

Prepare the model in the local directory $MODEL_PATH/{model_name}:

# Using Huggingface
huggingface-cli download {model_name} --local-dir $MODEL_PATH/{model_name}

# Using Modelscope
modelscope download {model_name} --local_dir $MODEL_PATH/{model_name}

For more details about model downloading, please refer to Huggingface or ModelScope.

Prepare the dataset in the local directory $DATASET_PATH/{dataset_name}:

# Using Huggingface
huggingface-cli download {dataset_name} --repo-type dataset --local-dir $DATASET_PATH/{dataset_name}

# Using Modelscope
modelscope download --dataset {dataset_name} --local_dir $DATASET_PATH/{dataset_name}

For more details about dataset downloading, please refer to Huggingface or ModelScope.

Step 3: configurations

For convenience, Trinity-RFT provides a web interface for configuring your RFT process.

Note

This is an experimental feature, and we will continue to improve it.

To enable minimal features (mainly for trainer), you can run

trinity studio --port 8080

Then you can configure your RFT process in the web page and generate a config file. You can save the config for later use or run it directly as described in the following section.

Advanced users can also configure the RFT process by editing the config file directly. We provide a set of example config files in examples.

To enable complete visualization features, please refer to the monorepo for Trinity-Studio.

Step 4: run the RFT process

First, start a ray cluster with the following command:

# On master node
ray start --head

# On worker nodes
ray start --address=<master_address>

Optionally, we can login into wandb to better monitor the RFT process:

export WANDB_API_KEY=<your_api_key>
wandb login

Then, for command-line users, run the RFT process with the following command:

trinity run --config <config_path>

For example, below is the command for fine-tuning Qwen-2.5-1.5B-Instruct on GSM8k dataset using GRPO algorithm:

trinity run --config examples/grpo_gsm8k/gsm8k.yaml

For studio users, just click the "Run" button in the web page.

For more detailed examples about how to use Trinity-RFT, please refer to the following tutorials:

• A quick example with GSM8k
• Off-policy mode of RFT
• Asynchronous mode of RFT
• Multi-turn tasks
• Offline learning by DPO
• Advanced data processing / human-in-the-loop

Advanced usage and full configurations

Please refer to this document.

Programming guide for developers

Please refer to this document.

Upcoming features

A tentative roadmap: #51

Contribution guide

This project is currently under active development, and we welcome contributions from the community!

Code style check:

pre-commit run --all-files

Unit tests:

python -m pytest tests

Acknowledgements

This project is built upon many excellent open-source projects, including:

• verl and PyTorch's FSDP for LLM training;
• vLLM for LLM inference;
• Data-Juicer for data processing pipelines;
• AgentScope for agentic workflow;
• Ray for distributed systems;
• we have also drawn inspirations from RL frameworks like OpenRLHF, TRL and ChatLearn;
• ......

Citation

@misc{trinity-rft,
      title={Trinity-RFT: A General-Purpose and Unified Framework for Reinforcement Fine-Tuning of Large Language Models},
      author={Xuchen Pan and Yanxi Chen and Yushuo Chen and Yuchang Sun and Daoyuan Chen and Wenhao Zhang and Yuexiang Xie and Yilun Huang and Yilei Zhang and Dawei Gao and Yaliang Li and Bolin Ding and Jingren Zhou},
      year={2025},
      eprint={2505.17826},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2505.17826},
}