apple/ml-fastvit
FastViT is a hybrid vision transformer architecture for image classification that uses structural reparameterization to achieve efficient inference on mobile devices.
Not currently ranked — collecting fresh signals.
star history
This repository provides the official PyTorch implementation of FastViT, a hybrid vision transformer that combines convolutional and transformer layers for image classification. The model uses structural reparameterization to convert multi-branch training blocks into efficient single-branch inference blocks. All models are trained on ImageNet-1K and benchmarked on mobile devices including iPhone 12 Pro. The implementation includes training code, pre-trained checkpoints, and CoreML models for deployment.
Frequently asked
- What is apple/ml-fastvit?
- FastViT is a hybrid vision transformer architecture for image classification that uses structural reparameterization to achieve efficient inference on mobile devices.
- Is ml-fastvit open source?
- Yes — apple/ml-fastvit is an open-source project tracked on heatdrop.
- What language is ml-fastvit written in?
- apple/ml-fastvit is primarily written in Python.
- How popular is ml-fastvit?
- apple/ml-fastvit has 2k stars on GitHub.
- Where can I find ml-fastvit?
- apple/ml-fastvit is on GitHub at https://github.com/apple/ml-fastvit.