veRL is a flexible and efficient RL framework for LLMs.

A simple, performant and scalable Jax LLM!

DeepSpeed version of NVIDIA's Megatron-LM that adds additional support for several features such as MoE model training, Curriculum Learning, 3D Parallelism, and others.

Efficient Training for Big Models.

Generative AI framework built for researchers and PyTorch developers working on Large Language Models (LLMs), Multimodal Models (MMs), Automatic Speech Recognition (ASR), Text to Speech (TTS), and Computer Vision (CV) domains.

Ongoing research training transformer models at scale.