nano-verl
A lightweight verl-style RL training framework.
File Explorer
Download Latest Version (.zip)- gsm8k-qwen3-0.6b-single-gpu.yaml
- gsm8k-qwen3-1.7b-1p1-async.yaml
- test.parquet
- train.parquet
- logo.png
- reward.png
- __init__.py
- actor_manager.py
- fsdp_worker.py
- __init__.py
- parquet_dataset.py
- __init__.py
- reward.py
- reward_manager.py
- __init__.py
- load_balancer.py
- rollout_manager.py
- vllm_server.py
- __init__.py
- base_engine.py
- naive_sync_engine.py
- nccl_sync_engine.py
- sync_manager.py
- weight_payload.py
- __init__.py
- trainer.py
- __init__.py
- algorithms.py
- config.py
- data.py
- logging.py
- main.py
- .gitignore
- main.py
- pyproject.toml
- README.md
- README_CN.md
# Installation Guide
1. Get the code
git clone https://github.com/kidding-404/nano-verl
Downloads the entire project code from GitHub to your computer.
cd nano-verl
Moves into the project folder you just downloaded.
2. Python
Easy RecommendedPrerequisites
uv run pip install <flash_attn_wheel_url>
Installs the Python libraries listed in requirements.txt (or similar).
If it runs without errors and prints output in the terminal, it worked.
Pulled directly from this repo's README.
// repository documentation
Was this content helpful?
(0 ratings)