cosmos-rl 0.4.3

Cosmos-RL is a flexible and scalable Reinforcement Learning framework specialized for Physical AI applications.

Getting Started

Cosmos-RL is a flexible and scalable Reinforcement Learning framework specialized for Physical AI applications.

Documentation.

System Architecture

Cosmos-RL provides toolchain to enable large scale RL training workload with following features:

1. Parallelism
   • Tensor Parallelism
   • Sequence Parallelism
   • Context Parallelism
   • FSDP Parallelism
   • Pipeline Parallelism
2. Fully asynchronous (replicas specialization)
   • Policy (Consumer): Replicas of training instances
   • Rollout (Producer): Replicas of generation engines
   • Low-precision training (FP8) and rollout (FP8 & FP4) support
3. Single-Controller Architecture
   • Efficient messaging system (e.g., weight-sync, rollout, evaluate) to coordinate policy and rollout replicas
   • Dynamic NCCL Process Groups for on-the-fly GPU [un]registration to enable fault-tolerant and elastic large-scale RL training

License and Contact

This project will download and install additional third-party open source software projects. Review the license terms of these open source projects before use.

NVIDIA Cosmos source code is released under the Apache 2 License.

NVIDIA Cosmos models are released under the NVIDIA Open Model License. For a custom license, please contact [e-mail].