GrowTAS: Progressive Expansion from Small to Large Subnets for Efficient ViT Architecture Search

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lee, Hyunju, Oh, Youngmin, Jeon, Jeimin, Baek, Donghyeon, Ham, Bumsub
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912762121158656
author Lee, Hyunju
Oh, Youngmin
Jeon, Jeimin
Baek, Donghyeon
Ham, Bumsub
author_facet Lee, Hyunju
Oh, Youngmin
Jeon, Jeimin
Baek, Donghyeon
Ham, Bumsub
contents Transformer architecture search (TAS) aims to automatically discover efficient vision transformers (ViTs), reducing the need for manual design. Existing TAS methods typically train an over-parameterized network (i.e., a supernet) that encompasses all candidate architectures (i.e., subnets). However, all subnets share the same set of weights, which leads to interference that degrades the smaller subnets severely. We have found that well-trained small subnets can serve as a good foundation for training larger ones. Motivated by this, we propose a progressive training framework, dubbed GrowTAS, that begins with training small subnets and incorporate larger ones gradually. This enables reducing the interference and stabilizing a training process. We also introduce GrowTAS+ that fine-tunes a subset of weights only to further enhance the performance of large subnets. Extensive experiments on ImageNet and several transfer learning benchmarks, including CIFAR-10/100, Flowers, CARS, and INAT-19, demonstrate the effectiveness of our approach over current TAS methods
format Preprint
id arxiv_https___arxiv_org_abs_2512_12296
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GrowTAS: Progressive Expansion from Small to Large Subnets for Efficient ViT Architecture Search
Lee, Hyunju
Oh, Youngmin
Jeon, Jeimin
Baek, Donghyeon
Ham, Bumsub
Computer Vision and Pattern Recognition
Machine Learning
Transformer architecture search (TAS) aims to automatically discover efficient vision transformers (ViTs), reducing the need for manual design. Existing TAS methods typically train an over-parameterized network (i.e., a supernet) that encompasses all candidate architectures (i.e., subnets). However, all subnets share the same set of weights, which leads to interference that degrades the smaller subnets severely. We have found that well-trained small subnets can serve as a good foundation for training larger ones. Motivated by this, we propose a progressive training framework, dubbed GrowTAS, that begins with training small subnets and incorporate larger ones gradually. This enables reducing the interference and stabilizing a training process. We also introduce GrowTAS+ that fine-tunes a subset of weights only to further enhance the performance of large subnets. Extensive experiments on ImageNet and several transfer learning benchmarks, including CIFAR-10/100, Flowers, CARS, and INAT-19, demonstrate the effectiveness of our approach over current TAS methods
title GrowTAS: Progressive Expansion from Small to Large Subnets for Efficient ViT Architecture Search
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2512.12296