# Block Selective Reprogramming for On-device Training of Vision Transformers

arXiv: https://arxiv.org/abs/2405.10951 · HTML: https://mlsanity.com/paper/2405.10951/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Vision mixture-of-experts, one of 12 directions in Mixture-of-experts.
- Builds on 32 of its 48 references that are in the map.
- First posted on arXiv 25 Mar 2024.

## Placed in

- Mixture-of-experts / [Vision mixture-of-experts](https://mlsanity.com/moe/vision-mixture-of-experts/)

## Builds on (most cited first, 30 of 32)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift](https://mlsanity.com/paper/1502.03167/) (2015)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding](https://mlsanity.com/paper/1510.00149/) (2015)
- [The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks](https://mlsanity.com/paper/1803.03635/) (2018)
- [Learning both Weights and Connections for Efficient Neural Networks](https://mlsanity.com/paper/1506.02626/) (2015)
- [Rethinking Attention with Performers](https://mlsanity.com/paper/2009.14794/) (2020)
- [Linformer: Self-Attention with Linear Complexity](https://mlsanity.com/paper/2006.04768/) (2020)
- [BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models](https://mlsanity.com/paper/2106.10199/) (2021)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [Training Deep Nets with Sublinear Memory Cost](https://mlsanity.com/paper/1604.06174/) (2016)
- [Channel Pruning for Accelerating Very Deep Neural Networks](https://mlsanity.com/paper/1707.06168/) (2017)
- [BinaryConnect: Training Deep Neural Networks with binary weights during propagations](https://mlsanity.com/paper/1511.00363/) (2015)
- [DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition](https://mlsanity.com/paper/1310.1531/) (2013)
- [Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations](https://mlsanity.com/paper/2202.07800/) (2022)
- [CNN Features off-the-shelf: an Astounding Baseline for Recognition](https://mlsanity.com/paper/1403.6382/) (2014)
- [EfficientFormer: Vision Transformers at MobileNet Speed](https://mlsanity.com/paper/2206.01191/) (2022)
- [Mobile-Former: Bridging MobileNet and Transformer](https://mlsanity.com/paper/2108.05895/) (2021)
- [Return of the Devil in the Details: Delving Deep into Convolutional Nets](https://mlsanity.com/paper/1405.3531/) (2014)
- [M3ViT: Mixture-of-Experts Vision Transformer for Efficient Multi-task Learning with Model-Accelerator Co-design](https://mlsanity.com/paper/2210.14793/) (2022)
- [Unified Visual Transformer Compression](https://mlsanity.com/paper/2203.08243/) (2022)
- [Memory-Efficient Backpropagation Through Time](https://mlsanity.com/paper/1606.03401/) (2016)
- [Sparse MoE as the New Dropout: Scaling Dense and Self-Slimmable Transformers](https://mlsanity.com/paper/2303.01610/) (2023)
- [Fusing Models with Complementary Expertise](https://mlsanity.com/paper/2310.01542/) (2023)
- [A Tunable Robust Pruning Framework Through Dynamic Network Rewiring of DNNs](https://mlsanity.com/paper/2011.03083/) (2020)
- [Efficient Knowledge Distillation from Model Checkpoints](https://mlsanity.com/paper/2210.06458/) (2022)
- [Learning to Linearize Deep Neural Networks for Secure and Efficient Private Inference](https://mlsanity.com/paper/2301.09254/) (2023)
- [Backprop with Approximate Activations for Memory-efficient Network Training](https://mlsanity.com/paper/1901.07988/) (2019)
- [Mesa: A Memory-saving Training Framework for Transformers](https://mlsanity.com/paper/2111.11124/) (2021)
- [Pre-defined Sparsity for Low-Complexity Convolutional Neural Networks](https://mlsanity.com/paper/2001.10710/) (2020)

## Built on by (most cited first, 4 of 4)

- [Adaptive Layer Selection for Efficient Vision Transformer Fine-Tuning](https://mlsanity.com/paper/2408.08670/) (2024)
- [MaskVD: Region Masking for Efficient Video Object Detection](https://mlsanity.com/paper/2407.12067/) (2024)
- [CHIPS: Efficient CLIP Adaptation via Curvature-aware Hybrid Influence-based Data Selection](https://mlsanity.com/paper/2511.18519/) (2025)
- [Decoupled Multi-Predictor Optimization for Inference-Efficient Model Tuning](https://mlsanity.com/paper/2511.03245/) (2025)

## Abstract (arXiv)

The ubiquity of vision transformers (ViTs) for various edge applications, including personalized learning, has created the demand for on-device fine-tuning. However, training with the limited memory and computation power of edge devices remains a significant challenge. In particular, the memory required for training is much higher than that needed for inference, primarily due to the need to store activations across all layers in order to compute the gradients needed for weight updates. Previous works have explored reducing this memory requirement via frozen-weight training as well storing the activations in a compressed format. However, these methods are deemed inefficient due to their inability to provide training or inference speedup. In this paper, we first investigate the limitations of existing on-device training methods aimed at reducing memory and compute requirements. We then present block selective reprogramming (BSR) in which we fine-tune only a fraction of total blocks of a pre-trained model and selectively drop tokens based on self-attention scores of the frozen layers. To show the efficacy of BSR, we present extensive evaluations on ViT-B and DeiT-S with five different datasets. Compared to the existing alternatives, our approach simultaneously reduces training memory by up to 1.4x and compute cost by up to 2x while maintaining similar accuracy. We also showcase results for Mixture-of-Expert (MoE) models, demonstrating the effectiveness of our approach in multitask learning scenarios.