PyPI page
Home page
Summary:
Cosmos-RL is a flexible and scalable Reinforcement Learning framework specialized for Physical AI applications.
Latest version:
0.4.6
Required dependencies:
accelerate
|
blobfile
|
boto3
|
click
|
cloudpickle
|
datasets
|
diffusers
|
fastapi
|
kernels
|
liger-kernel
|
math-verify
|
modelscope
|
msgpack
|
nest-asyncio
|
nvidia-ml-py
|
nvidia-nccl-cu12
|
peft
|
pybind11
|
pydantic
|
pynvml
|
pyyaml
|
qwen-vl-utils
|
redis
|
rich
|
strenum
|
tensordict
|
toml
|
torch
|
torchao
|
torchvision
|
transformers
|
uvicorn
|
zmq
Optional dependencies:
aioboto3
|
aiobotocore
|
bddl
|
better-profanity
|
cosmos-rl
|
decord
|
easydict
|
einops
|
flash-attn
|
future
|
gym
|
gymnasium
|
hydra-core
|
imageio
|
imageio-ffmpeg
|
libero
|
math-verify
|
matplotlib
|
nltk
|
numpy
|
opencv-python
|
opencv-python-headless
|
retinaface-py
|
robomimic
|
robosuite
|
sentencepiece
|
termcolor
|
thop
|
timm
|
transformers
|
trimesh
|
ucxx-cu12
|
vllm
|
wandb
|
webdataset
Downloads last day:
1
Downloads last week:
64
Downloads last month:
294
7-day smoothing