The nnsight package enables interpreting and manipulating the internals of deep learned models.

Inference algorithms for models based on Luce's choice axiom

Steering vectors for transformer language models in Pytorch / Huggingface

Code for the TinyStories experiments from "Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks".

This is code for most of the experiments in the paper Understanding the Effects of RLHF on LLM Generalisation and Diversity

A browser extension that deletes your news feed and replaces it with a nice quote

ChatArena (or Chat Arena) is a Multi-Agent Language Game Environments for LLMs. The goal is to develop communication and collaboration capabilities of AIs.

lesspipe - display more with less

Adds vim keybindings to all OS X inputs

A library for mechanistic interpretability of GPT-style language models

🎢 Creating and sharing simulation environments for embodied and synthetic data research

A modular RL library to fine-tune language models to human preferences

A repo for distributed training of language models with Reinforcement Learning via Human Feedback (RLHF)

BertViz: Visualize Attention in NLP Models (BERT, GPT2, BART, etc.)

Mechanistic Interpretability for Transformer Models

Code for the paper Fine-Tuning Language Models from Human Preferences

Human preference data for "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback"

Plug and play RAM percentage and icon indicator for Tmux

An interactive NVIDIA-GPU process viewer and beyond, the one-stop solution for GPU process management.

Use Google Docs like vim. Sorta.

Train transformer language models with reinforcement learning.

A library for distributed ML training with PyTorch

[NeurIPS'21 Outstanding Paper] Library for reliable evaluation on RL and ML benchmarks, even with only a handful of seeds.

DMControl Generalization Benchmark

arXiv LaTeX Cleaner: Easily clean the LaTeX code of your paper to submit to arXiv

MiniHack the Planet: A Sandbox for Open-Ended Reinforcement Learning Research

PAIRED in PyTorch 🔥

A simple tool to update bib entries with their official information (e.g., DBLP or the ACL anthology).