mosaicml/composer
View on GitHub ↗ ★ 5,506 Python archived
Supercharge Your Model Training
8 mentions on Hacker News by 6 people: 4 stories, 4 comments · first 2022-03-16 · latest 2023-08-18
20082026
-
▲ 1
Composer – A PyTorch Library for Efficient Neural Network Trainingstory by SiempreViernes · 2023-08-18 · 0 comments
-
▲ 2
Train neural networks up to 7x fasterstory by udev4096 · 2023-06-30 · 0 comments
-
Mosaic's open source library is excellent: Composer https://github.com/mosaicml/composer . * It gives you PyTorch DDP for free. Makes FSDP about as easy as can be, and provides best in class performance monitoring tools.…
comment by NerdyDrone · 2023-02-11 · in How to train large models on many GPUs? (2021) -
…but I could definitely meet and talk through it, help code tricky parts, review PRs, etc. [1] https://github.com/mosaicml/composer
comment by ffast-math · 2022-06-19 · in Bolt: Faster matrix and vector operations that run on compressed data -
…and, in contrast to Bolt/MADDNESS, is polished, documented code you can get working in <5min. [1] https://github.com/mosaicml/composer [2] https://www.mosaicml.com/blog/mosaic-resnet
comment by ffast-math · 2022-06-19 · in Bolt: Faster matrix and vector operations that run on compressed data -
…should be a big source of efficiency. We've released a new PyTorch library for efficient training at http://github.com/mosaicml/composer . Our combinations of methods can train CV models ~4x faster to the same accuracy on CV tasks, and ~2x…
comment by moinnadeem · 2022-03-22 · in Nvidia Hopper GPU Architecture and H100 Accelerator -
▲ 3
Train ML Models 2-4x Faster with SOTA Researchstory by averylamp · 2022-03-17 · 0 comments
-
▲ 21
Composer: A new PyTorch library to train models 4x faster with better algorithmsstory by moinnadeem · 2022-03-16 · 0 comments