KellerJordan / Muon
Muon optimizer: +~30% sample efficiency with <3% wallclock overhead
☆251Updated last week
Alternatives and similar repositories for Muon:
Users that are interested in Muon are comparing it to the libraries listed below
- supporting pytorch FSDP for optimizers☆76Updated 2 months ago
- Efficient optimizers☆169Updated this week
- Understand and test language model architectures on synthetic tasks.☆181Updated last month
- Code for exploring Based models from "Simple linear attention language models balance the recall-throughput tradeoff"☆220Updated 2 months ago
- Normalized Transformer (nGPT)☆150Updated 2 months ago
- ☆157Updated 2 months ago
- Quick implementation of nGPT, learning entirely on the hypersphere, from NvidiaAI☆271Updated 3 months ago
- Repo for "LoLCATs: On Low-Rank Linearizing of Large Language Models"☆215Updated 2 weeks ago
- Pytorch implementation of the PEER block from the paper, Mixture of A Million Experts, by Xu Owen He at Deepmind☆117Updated 5 months ago
- When it comes to optimizers, it's always better to be safe than sorry☆179Updated 3 weeks ago
- ☆180Updated this week
- Some preliminary explorations of Mamba's context scaling.☆213Updated last year
- Accelerated First Order Parallel Associative Scan☆171Updated 5 months ago
- Helpful tools and examples for working with flex-attention☆630Updated this week
- Code for Adam-mini: Use Fewer Learning Rates To Gain More https://arxiv.org/abs/2406.16793☆385Updated 2 months ago
- ☆251Updated 5 months ago
- The simplest, fastest repository for training/finetuning medium-sized GPTs.☆94Updated 2 months ago
- WIP☆93Updated 6 months ago
- A MAD laboratory to improve AI architecture designs 🧪☆102Updated last month
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆114Updated 2 months ago
- PyTorch implementation of models from the Zamba2 series.☆176Updated 3 weeks ago
- DeMo: Decoupled Momentum Optimization☆180Updated 2 months ago
- Code to reproduce "Transformers Can Do Arithmetic with the Right Embeddings", McLeish et al (NeurIPS 2024)☆183Updated 8 months ago
- 🧱 Modula software package☆139Updated this week
- ☆75Updated 7 months ago
- Memory layers use a trainable key-value lookup mechanism to add extra parameters to a model without increasing FLOPs. Conceptually, spars…☆294Updated 2 months ago
- Implementation of 💍 Ring Attention, from Liu et al. at Berkeley AI, in Pytorch☆501Updated 3 months ago
- A subset of PyTorch's neural network modules, written in Python using OpenAI's Triton.☆513Updated this week
- seqax = sequence modeling + JAX☆142Updated 6 months ago
- The AdEMAMix Optimizer: Better, Faster, Older.☆177Updated 5 months ago