Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

Maximally Scale-Stable Parameterization (MSSP) for Mixture-of-Experts: hyperparameter transfer and predictable scaling across width, depth, expert count, and expert width.

Solve puzzles. Improve your pytorch.

The open source coding agent.

Serving JS to the browser

Measuring how well CLI agents like Claude Code or Codex CLI can post-train base LLMs on a single H100 GPU in 10 hours

Autoresearch for LLM adversarial attacks

MoE training for Me and You and maybe other people

Automatic tiling window manager for macOS à la xmonad.

An extension of the nanoGPT repository for training small MOE models.

Continual pretrainig experiments with large language models

Ongoing research training transformer models at scale

AllenAI's post-training codebase

Matplotlib styles for scientific plotting

Code for the Paper: "STAMP Your Content: Proving Dataset Membership via Watermarked Rephrasings"

Code release to accompany the paper "Persistent Pre-training Poisoning of LLMs"

[NeurIPS D&B '25] The one-stop repository for LLM unlearning

Open-source framework for the research and development of foundation models.

PyTorch building blocks for the OLMo ecosystem

Reproducible, flexible LLM evaluations

Monitor the training of your PyTorch models, including transformers

The code for creating the iGSM datasets in papers "Physics of Language Models Part 2.1, Grade-School Math and the Hidden Reasoning Process" (arxiv 2407.20311) and "Physics of Language Models Part 2…

A lightweight suffix-sorting library

What's In My Big Data (WIMBD) - a toolkit for analyzing large text datasets

View model summaries in PyTorch!

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

A library for unit scaling in PyTorch