Code for paper "Multiple Physics Pretraining for Physical Surrogate Models

Official implementation of "Lost in Latent Space: An Empirical Study of Latent Diffusion Models for Physics Emulation"

Code for "Physics Steering: Causal Control of Cross-Domain Concepts in a Physics Foundation Model" paper.

Cosmos-Transfer2.5, built on top of Cosmos-Predict2.5, produces high-quality world simulations conditioned on multiple spatial control inputs.

Polymathic's Large Omnimodal Model for Astronomy

The best ChatGPT that $100 can buy.

Cosmos-Predict2.5, the latest version of the Cosmos World Foundation Models (WFMs) family, specialized for simulating and predicting the future state of the world in the form of video.

Post-training scripts and samples for NVIDIA Cosmos ecosystem

A minimal implementation of DeepMind's Genie world model

3D Object Reconstruction project is a workflow that takes a set of stereo images and camera info and outputs a textured mesh (i.e., .OBJ file). The purpose is to translate physical items into the d…

Official implementation for the paper: InterTrack

codebase for iccv 2025 paper "One Trajectory, One Token: Grounded Video Tokenization via Panoptic Sub-object Trajectory"

Efficient optimizers

A 15TB Collection of Physics Simulation Datasets

MMPS Benchmark

Schedule-Free Optimization in PyTorch

General Relativistic Neutrino Radiation Magnetohydrodynamics for Neutron Star Merger Disks

Design and analyze optimal deep learning models.

VideoSys: An easy and efficient system for video generation

Code of the paper "Listening to the Noise: Blind Denoising with Gibbs Diffusion"

Image Restoration Toolbox (PyTorch). Training and testing codes for DPIR, USRNet, DnCNN, FFDNet, SRMD, DPSR, BSRGAN, SwinIR

OceanBench - SSH edition

Simple, minimal implementation of the Mamba SSM in one file of PyTorch.

Machine Learning Engineering Open Book

Fast & Simple repository for pre-training and fine-tuning T5-style models

Revisiting Efficient Training Algorithms For Transformer-based Language Models (NeurIPS 2023)

MoMo: Momentum Models for Adaptive Learning Rates

The simplest, fastest repository for training/finetuning medium-sized GPTs.

Official repo for consistency models.