NeMo RL
NVIDIA Apache-2.0 post-training / reinforcement-learning library for large-scale model training, with official docs and NGC containers.
Tags
Description
Established NVIDIA project (canonical repo NVIDIA-NeMo/RL; not archived; created 2025-03-16; not a 2026 launch). Older GitHub paths NVIDIA/NeMo-RL and NVIDIA/nemo-rl redirect to NVIDIA-NeMo/RL. Root LICENSE is Apache-2.0. GitHub latest release v0.8.0 was published 2026-11-06T22:22:08Z (re-confirmed 2026-11-13). PyPI package nemo-rl is still 0.0.0 (uploaded 2025-06-03, homepage still lists github.com/NVIDIA/NeMo-RL) and is not treated as the current install path. Official README quick start uses the NGC container nvcr.io/nvidia/nemo-rl:latest plus a git clone of NVIDIA-NeMo/RL, or a bare-metal setup described in the installation guide. Docs: https://docs.nvidia.com/nemo/rl/latest/index.html and the installation guide at https://docs.nvidia.com/nemo/rl/latest/about/installation.html (both HTTP 200). Training uses GPUs and may pull gated model weights; those costs and any cloud or NGC access are separate. This library can run training jobs that execute model and environment code — follow upstream safety guidance. Vendor: NVIDIA.
Agent access
Related
- NeMo Gym: NVIDIA Apache-2.0 library for building, evaluating and scaling agent environments, verifiers and harnesses, including sandboxed code-execution tasks.
- TRL: Hugging Face Transformer Reinforcement Learning library for post-training methods such as SFT, DPO, PPO and GRPO on top of transformers.
- verl: Distributed reinforcement-learning training library for LLMs, with GRPO/PPO dataflows and pluggable training and rollout engines for agentic post-training.
- SkyRL: Modular reinforcement-learning stack with a training engine, long-horizon agent layer and Gymnasium-style tool-use environments for LLMs.
- OpenRLHF: Apache-2.0 high-performance RLHF / post-training framework built on Ray, with support for multiple reward and policy-optimization recipes.
Guides & comparisons
- RL environment libraries compared — NeMo Gym, OpenEnv and Verifiers (Comparison): Compare NeMo Gym, OpenEnv and Verifiers as libraries for writing and running RL environments: authoring model, isolation, documented trainers, install paths, licences and current versions. Distinct from Harbor/AgentEnv/Seahaven and from the agent RL stacks comparison.
Sources
- https://github.com/NVIDIA-NeMo/RL
- https://github.com/NVIDIA-NeMo/RL/blob/main/LICENSE
- https://github.com/NVIDIA-NeMo/RL/blob/main/README.md
- https://github.com/NVIDIA-NeMo/RL/releases/tag/v0.8.0
- https://docs.nvidia.com/nemo/rl/latest/index.html
- https://docs.nvidia.com/nemo/rl/latest/about/installation.html
- https://pypi.org/project/nemo-rl/
Machine-readable
- JSON
- Markdown
- Source file on GitHub (edit via pull request)