verl

PyPI page
Home page
Author: Bytedance - Seed - MLSys
License: Apache-2.0
Summary: verl: Volcano Engine Reinforcement Learning for LLM
Latest version: 0.9.1
Required dependencies: accelerate | codetiming | datasets | dill | hydra-core | numpy | packaging | pandas | peft | pyarrow | pybind11 | pylatexenc | ray | tensorboard | tensordict | torchdata | transformers | wandb
Optional dependencies: accelerate | apex | cachetools | codetiming | cupy-cuda13x | datasets | dill | fastapi | flash-attn | flash-linear-attention | hf-transfer | hydra-core | liger-kernel | math-verify | mathruler | matplotlib | mbridge | megatron-bridge | megatron-core | msgspec | numpy | nvidia-mathdx | nvidia-modelopt | nvtx | onnxscript | packaging | pandas | peft | pre-commit | py-spy | pyarrow | pybind11 | pylatexenc | pytest | pytest-asyncio | pytest-rerunfailures | pyzmq | qwen-vl-utils | ray | sglang | sglang-kernel | sglang-router | tensorboard | tensordict | torch | torchaudio | torchcodec | torchdata | torchvision | transferqueue | transformer-engine | transformers | triton | trl | uvicorn | veomni | verl | vllm | wandb

Downloads last day: 758
Downloads last week: 4,655
Downloads last month: 26,607
7-day smoothing