fastai v2.7 Released
Version 2.7 of fastai has been released. This update brings Hugging Face Accelerate in to handle distributed training, among other nice changes.
Created on June 21|Last edited on June 21
Comment
fastai, created by the people at fast.ai, is an open-source Python library built to make deep learning more accessible, with loads of built-in functions that'll give people interesting in machine learning but without any background the ability to create the projects they want to - with just a little bit of Python understanding required.
Version 2.7 of fastai has been officially released (2.7.2 to be specific, as there were a couple quick bug fix patches).
What's new in fastai v2.7?
The biggest change of note in fastai v2.7 is that distributed training is now handled by Hugging Face Accelerate. Thanks to this big change, distributed training is now also supported in notebooks.
Other changes are not as groundbreaking, being mostly tweaks, additional support, or bug fixes. A few things that might be good to know are that a workaround has been made for a performance bug in PyTorch regarding subclassing, as well as fastai now being compatible with PyTorch 1.12 (the current preview version of PyTorch).
For the full list of changes you can take a look at the github changelog here: https://github.com/fastai/fastai/releases/tag/2.7.0
Find out more
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.