Agentic RL Training at Scale

100M tokens. Infinite compute. Lowest val loss wins.

Connect any Open Data to any LLM with Model Context Protocol.

a repo for some beancount tool

Importers framework for Beancount

Beancount: Double-Entry Accounting from Text Files.

Web UI for semi-automatically importing external data into beancount

Visualize, query, and stream to train on multimodal robotics data.

JAX Synergistic Memory Inspector

Important concepts in numerical linear algebra and related areas

Elegant easy-to-use neural networks + scientific computing in JAX. https://docs.kidger.site/equinox/

Transformer related optimization, including BERT, GPT

magic-trace collects and displays high-resolution traces of what a process is doing

Containers for machine learning

Memray is a memory profiler for Python

Lightning fast C++/CUDA neural network framework

nanobind: tiny and efficient C++/Python bindings

a structural diff that understands syntax 🟥🟩

Context-based code search tool

get things from one computer to another, safely

A pattern-based approach to learn technical interview questions

An attempt to answer the age old interview question "What happens when you type google.com into your browser and press enter?"

Pytorch implementation of MaskGIT: Masked Generative Image Transformer (https://arxiv.org/pdf/2202.04200.pdf)

A retargetable MLIR-based machine learning compiler and runtime toolkit.

Algorithms for outlier, adversarial and drift detection

ML Collections is a library of Python Collections designed for ML use cases.

Library for 8-bit optimizers and quantization routines.

Connect to Google Colab using SSH

neural image generation

Jupyter notebooks for the Natural Language Processing with Transformers book