Super basic implementation (gist-like) of RLMs with REPL environments.

"ViMax: Agentic Video Generation (Director, Screenwriter, Producer, and Video Generator All-in-One)"

nanoGRPO is a lightweight implementation of Group Relative Policy Optimization (GRPO)

A prototype implementation of the "dataset as a queue" pattern for processing web pages into interleaved image/text content.

Ongoing research training transformer models at scale

🚀 Pytorch Distributed native training library for LLMs/VLMs with OOTB Hugging Face support

Multilingual Document Layout Parsing in a Single Vision-Language Model

This repo contains the code for 1D tokenizer and generator

This is the official repo for the paper "LongCat-Flash-Omni Technical Report"

Simple IO APIs with pluggable storage backends and rich format handlers.

Official Implementation of the ICCV 2023 paper: Perpetual Humanoid Control for Real-time Simulated Avatars

Tiny AutoEncoder for Hunyuan Video (and other video models)

Cosmos-RL is a flexible and scalable Reinforcement Learning framework specialized for Physical AI applications.

RES: Refined Exponential Solver. https://arxiv.org/abs/2308.02157

Official repo for paper "MiraData: A Large-Scale Video Dataset with Long Durations and Structured Captions"

[NeurIPS 2025] Radial Attention: O(nlogn) Sparse Attention with Energy Decay for Long Video Generation

[ICML2025, NeurIPS2025 Spotlight] Sparse VideoGen 1 & 2: Accelerating Video Diffusion Transformers with Sparse Attention

Best practices for training DeepSeek, Mixtral, Qwen and other MoE models using Megatron Core.

Dion optimizer algorithm

Evaluation harness for diffusion world models

A place to store reusable transformer components of my own creation or found on the interwebs

Kernels, of the mega variety :)

Cosmos-Predict2 is a collection of general-purpose world foundation models for Physical AI that can be fine-tuned into customized world models for downstream applications.

open-source coding LLM for software engineering tasks

[ICML 2025 Spotlight] Direct Discriminative Optimization: Reinforcing Diffusion/Autoregressive with GAN Discrimination

The official implementation of CVPR'25 Oral paper "Go-with-the-Flow: Motion-Controllable Video Diffusion Models Using Real-Time Warped Noise"

Quickly rewrite git repository history (filter-branch replacement)