⚡ Migrating from PTL - Composer
PyTorch Lightning is a popular and well-designed framework for training deep neural networks. You can use Composer’s algorithms in your Pytorch Lightning code via the functional API with no additio...
Tip This tutorial is available as a Jupyter notebook . ⚡ Migrating from PTL # PyTorch Lightning is a popular and well-designed framework for training deep neural networks. You can use Composer’s algorithms in your Pytorch Lightning code via the functional API with no additional code changes. However, if you are interested in features like automatic gradient accumulation , a clean time abstraction , and the easiest path to trying out different combinations of algorithms, you will need to switch from the PTL trainer to the Composer trainer. The below is a quick guide on how to adapt your Lightni
saved by
related reading
- ⏱️ Performance Profiling - Composerdocs.mosaicml.com
- 👨👩👧👦 Distributed Training - Composerdocs.mosaicml.com
- 🥽 Analyzing Tracesdocs.mosaicml.com
- ♻️ Auto Microbatching - Composerdocs.mosaicml.com
- ⏯️ Autoresume Training - Composerdocs.mosaicml.com
- Mosaic ResNet Deep Dive | Databricks Blogmosaicml.com
- LightningModule - PyTorch Lightning 2.6.1 documentationlightning.ai
- A friendly introduction to machine learning compilers and optimizershuyenchip.com
- Trainer - PyTorch Lightning 2.6.1 documentationlightning.ai
- Making Deep Learning go Brrrr From First Principleshorace.io
- Get Started with Distributed Training using PyTorch Lightning — Ray 2.56.0docs.ray.io
- UW PLSE | How does torch.compile work?uwplse.org