Features and Roadmap#
Available Now#
Distributed Training - Ray-based infrastructure.
Environment Support and Isolation - Support for multi-environment training and dependency isolation between components.
Worker Isolation - Process isolation between RL Actors (no worries about global state).
Learning Algorithms - GRPO/GSPO/DAPO, SFT (with LoRA), DPO, and on-policy distillation.
Multi-Turn RL - Multi-turn generation and training for RL with tool use, games, etc.
Advanced Parallelism with DTensor - PyTorch FSDP2, TP, CP, and SP for efficient training (through NeMo AutoModel).
Larger Model Support with Longer Sequences - Performant parallelism with Megatron Core (TP/PP/CP/SP/EP/FSDP) through NeMo Megatron Bridge.
Sequence Packing - Sequence packing in both DTensor and Megatron Core for large training performance gains.
Fast Generation - vLLM backend for optimized inference.
Hugging Face Integration - Out-of-box support in the DTensor path, with checkpoint conversion available for the Megatron path through Megatron Bridge middleware.
End-to-End FP8 Low-Precision Training - Support for Megatron Core FP8 training and FP8 vLLM generation.
Vision Language Models (VLM) - Support SFT and GRPO on VLMs.
Megatron Inference - Megatron Inference for fast day-0 support for new Megatron models without weight conversion.
Async RL - Support for asynchronous rollouts and replay buffers for off-policy training, and enable a fully asynchronous GRPO.
NeMo-Gym Integration - RL environment integration.
GB200 - Container support for GB200.
Planned / In Progress#
Muon Optimizer - Emerging optimizer support for SFT/RL.
SGLang Inference - SGLang rollout support for optimized inference.
Improved Native Performance - Improved training time for native PyTorch models.
Improved Large MoE Performance - Improved Megatron Core training and generation performance.
New Models - Qwen3-Next and Nemotron-Super.
Expanded Algorithms - GDPO and broader LoRA coverage for GRPO and DPO.
Resiliency - Fault tolerance and auto-scaling support.
Speculative Decoding - Speculative decoding support for rollout acceleration.