Python wrapper for the NVIDIA cuSFM library

VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo

Repo for SeedVR2 & SeedVR (CVPR2025 Highlight)

SeedVR2: One-Step Video Restoration via Diffusion Adversarial Post-Training

Reverse Engineering Gemma 3n: Google's New Edge-Optimized Language Model

[NeurIPS 2025] MMaDA - Open-Sourced Multimodal Large Diffusion Language Models

Open-source unified multimodal model

[CVPR 2025 Best Paper Award] VGGT: Visual Geometry Grounded Transformer

A bidirectional pipeline parallelism algorithm for computation-communication overlap in DeepSeek V3/R1 training.

Wan: Open and Advanced Large-Scale Video Generative Models

FlashMLA: Efficient Multi-head Latent Attention Kernels

Production-tested AI infrastructure tools for efficient AGI development and community-driven innovation

MoBA: Mixture of Block Attention for Long-Context LLMs

[CVPR 2025 Highlight] Video Depth Anything: Consistent Depth Estimation for Super-Long Videos

Vchitect-2.0: Parallel Transformer for Scaling Up Video Diffusion Models

Official Repo For "Sa2VA: Marrying SAM2 with LLaVA for Dense Grounded Understanding of Images and Videos"

New repo collection for NVIDIA Cosmos: https://github.com/nvidia-cosmos

[ICCV 2025] Edicho: Consistent Image Editing in the Wild

HunyuanVideo: A Systematic Framework For Large Video Generation Model

[CVPR'25] Official Implementations for Paper - MagicQuill: An Intelligent Interactive Image Editing System

AlphaFold 3 inference pipeline.

A suite of image and video neural tokenizers

Code for "Diffusion Model Alignment Using Direct Preference Optimization"

Video Generation, Physical Commonsense, Semantic Adherence, VideoCon-Physics

Scaling Diffusion Transformers with Mixture of Experts

Meta Lingua: a lean, efficient, and easy-to-hack codebase to research LLMs.

Official Implementation of paper "MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion"