[ICLR 2026] Official PyTorch Implementation of RLP: Reinforcement as a Pretraining Objective

Fast and easy to use CIFAR-10 dataloader

RWKV, in easy to read code

1.2% test error on MNIST using only least squares and numpy calls.

Python bindings for the Transformer models implemented in C/C++ using GGML library.

LLM inference in C/C++

Root Mean Square Layer Normalization

Minimalistic, extremely fast, and hackable researcher's toolbench for GPT models in 307 lines of code. Reaches <3.8 validation loss on wikitext-103 on a single A100 in <100 seconds. Scales to large…

A PyTorch library entirely dedicated to neural differential equations, implicit models and related numerical methods

Running large language models on a single GPU for throughput-oriented scenarios.

Stable diffusion for real-time music generation

A template for two-column scientific preprints

Physical modelling signal processing for analog tape recording

Customizable, site-specific searches directly from your browser's search input. Personal DuckDuckGo search !bangs.

Train CIFAR10 to 94% accuracy in a few minutes/seconds. Based on https://github.com/davidcpage/cifar10-fast

Never use print for debugging again

A cross-platform command-line utility that creates projects from cookiecutters (project templates), e.g. Python package projects, C projects.

Artist-friendly Modular Machine for Art and Technology

Train to 94% on CIFAR-10 in <6.3 seconds on a single A100. Or ~95.79% in ~110 seconds (or less!)