maximal update parametrization (µP)

Simulation benchmark from Toyota Research Institute containing 49 tasks that measure the performance of Large Behavior Model policies

DayDreamer: World Models for Physical Robot Learning

Running VLA at 30Hz frame rate and 480Hz trajectory frequency

Visual understanding of motion for policy learning

An open-source, GPU-accelerated physics simulation engine built upon NVIDIA Warp, specifically targeting roboticists and simulation researchers.

SERL: A Software Suite for Sample-Efficient Robotic Reinforcement Learning

Being-H0: Vision-Language-Action Pretraining from Large-Scale Human Videos

A collection of tabletop tasks in Mujoco

This is a library that provides dual dexterous hand manipulation tasks through Isaac Gym

Geometric Retargeting A Principled, Ultrafast Neural Hand Retargeting Algorithm

Unified framework for robot learning built on NVIDIA Isaac Sim

HumanML3D: A large and diverse 3d human motion-language dataset.

The official PyTorch implementation of the paper "Human Motion Diffusion Model"

Single File, Single GPU, From Scratch, Efficient, Full Parameter Tuning library for "RL for LLMs"

Line-by-line profiling for Python

Minimal, clean code for the Byte Pair Encoding (BPE) algorithm commonly used in LLM tokenization.

Muon is an optimizer for hidden layers in neural networks

End-to-End Navigation with VLMs

Bash Line Editor―a line editor written in pure Bash with syntax highlighting, auto suggestions, vim modes, etc. for Bash interactive sessions.

Official implementation of paper: SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training

A Telegram bot to recommend arXiv papers

[CVPR'23] Video Probabilistic Diffusion Models in Projected Latent Space