Stars
FDFO: Finite Difference Flow Optimization
[NeurIPS 2025] An official implementation of Flow-GRPO: Training Flow Matching Models via Online RL
Continuous Unix commit history from 1970 until today
Visual Imitation Enables Contextual Humanoid Control. CoRL 2025, Best Student Paper Award.
Simplified Perpetual Humanoid Control with Pufferlib, CARBS
NVIDIA Isaac GR00T N1.7 - A Foundation Model for Generalist Robots.
Official implementation of Inductive Moment Matching
Development repository for the Triton language and compiler
Minimal reproduction of DeepSeek R1-Zero
NVIDIA Cosmos is an open platform of world models, datasets, and tools that enables developers to build Physical AI for robots, autonomous vehicles, smart infrastructure, and more.
🚀 Efficient implementations for emerging model architectures
Adaptive Length Image Tokenization via Recurrent Allocation | How many tokens is an image worth ?
A suite of image and video neural tokenizers
Patch convolution to avoid large GPU memory usage of Conv2D
ElasticTok: Adaptive Tokenization for Image and Video
FlashTex: Fast Relightable Mesh Texturing with LightControlNet
Evaluating and reproducing real-world robot manipulation policies (e.g., RT-1, RT-1-X, Octo) in simulation under common setups (e.g., Google Robot, WidowX+Bridge) (CoRL 2024)
The simplest, fastest repository for training/finetuning medium-sized GPTs.
Official inference repo for FLUX.1 models
Ongoing research training transformer models at scale
Official Implementation of Rethinking Score Distillation as a Bridge Between Image Distributions
Evaluating text-to-image/video/3D models with VQAScore
[NeurIPS 2024]OmniTokenizer: one model and one weight for image-video joint tokenization.
[ICLR'24] Efficient Video Diffusion Models via Content-Frame Motion-Latent Decomposition
A framework for 4D reconstruction from monocular videos.
Code repo for "WebArena: A Realistic Web Environment for Building Autonomous Agents"