optimizer & lr scheduler & loss function collections in PyTorch

muon is a multimodal omics Python framework

Multimodal Data (.h5mu) implementation for Python

Multimodal datasets, in MuData format

Use of Deep Mixture Density Networks to predict Muon Hit positions from Photon propagation time in scintillator bars in Muon Telescopes

Simple muon position reconstruction algorithm in a plastic scintillator

Materials for the PUMAS library

A modern minimal LLM implementation made to be easily modified by non-professionals and trained on consumer hardware.

A university-based physics project measuring muon lifetime using detector geometry, Monte Carlo simulation, and Python visualization to estimate effective area and particle flux.

raw datasets corresponding to the findings presented in arXiv:2412.08349

This version of Muon converges slightly faster than the Muon from modded-nanogpt in some cases. The change is RMS-Norm after orthogonalization over the first dimension of the weight matrix (last dimension of nn.Linear). The code here assumes you store the weights like nn.Linear i.e. used like x = x @ W.T.

raw datasets corresponding to the findings presented in arXiv:2512.12481

🚀 Accelerate language model training with modded-nanogpt, achieving 3.28 loss in under 3 minutes on 8 NVIDIA H100 GPUs for faster AI development.

🔬 Explore and analyze multimodal omics data with the muon framework, designed for efficient handling of diverse biological datasets in Python.

🔍 Explore multimodal omics data easily with `muon`, a powerful Python framework designed for efficient analysis and visualization of diverse biological datasets.