A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.

A library of reinforcement learning components and agents

C++-based high-performance parallel environment execution engine (vectorized env) for general RL environments.

SEED RL: Scalable and Efficient Deep-RL with Accelerated Central Inference. Implements IMPALA and R2D2 algorithms in TF2 with SEED's architecture.

An elegant PyTorch deep reinforcement learning library.

Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.

Scalable toolkit for efficient model reinforcement

An early research stage expert-parallel load balancer for MoE models based on linear programming.

A training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.

PyTorch version of Stable Baselines, reliable implementations of reinforcement learning algorithms.

verl: Volcano Engine Reinforcement Learning for LLMs

Clean, Robust, and Unified PyTorch implementation of popular Deep Reinforcement Learning (DRL) algorithms (Q-learning, Duel DDQN, PER, C51, Noisy DQN, PPO, DDPG, TD3, SAC, ASL)

FAIR Sequence Modeling Toolkit 2

High-Performance Symbolic Regression in Python and Julia

A collection of AWESOME things about mixture-of-experts

Implementations of a Mixture-of-Experts (MoE) architecture designed for research on large language models (LLMs) and scalable neural network designs. One implementation targets a **single-device/NP…

DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models

A PyTorch native platform for training generative AI models

[ICCV 2025] Official implementation of the paper: REPA-E: Unlocking VAE for End-to-End Tuning of Latent Diffusion Transformers

Hackable and optimized Transformers building blocks, supporting a composable construction.

Best practice for training LLaMA models in Megatron-LM

A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)

Causal depthwise conv1d in CUDA, with a PyTorch interface

A Library for Advanced Deep Time Series Models for General Time Series Analysis.

🚀 Efficient implementations for emerging model architectures

A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training

MAGI-1: Autoregressive Video Generation at Scale

FlashRNN - Fast RNN Kernels with I/O Awareness

High-Resolution Image Synthesis with Latent Diffusion Models