MobileTL: On-device Transfer Learning with Inverted Residual Blocks

TL;DR

MobileTL enables efficient on-device transfer learning for models with Inverted Residual Blocks by reducing memory and computation costs, making transfer learning feasible on resource-limited edge devices.

Contribution

The paper introduces MobileTL, a novel method that allows transfer learning with IRB-based models on edge devices by approximating backward computations and fine-tuning top layers.

Findings

Reduces memory usage by 46% and 53% for MobileNetV2 and V3 IRBs.

Achieves 36% FLOPs reduction with minimal accuracy loss on CIFAR10.

Demonstrates Pareto-optimal performance across multiple datasets.

Abstract

Transfer learning on edge is challenging due to on-device limited resources. Existing work addresses this issue by training a subset of parameters or adding model patches. Developed with inference in mind, Inverted Residual Blocks (IRBs) split a convolutional layer into depthwise and pointwise convolutions, leading to more stacking layers, e.g., convolution, normalization, and activation layers. Though they are efficient for inference, IRBs require that additional activation maps are stored in memory for training weights for convolution layers and scales for normalization layers. As a result, their high memory cost prohibits training IRBs on resource-limited edge devices, and making them unsuitable in the context of transfer learning. To address this issue, we present MobileTL, a memory and computationally efficient on-device transfer learning method for models built with IRBs. MobileTL trains the shifts for internal normalization layers to avoid storing activation maps for the backward pass. Also, MobileTL approximates the backward computation of the activation layer (e.g., Hard-Swish and ReLU6) as a signed function which enables storing a binary mask instead of activation maps for the backward pass. MobileTL fine-tunes a few top blocks (close to output) rather than propagating the gradient through the whole network to reduce the computation cost. Our method reduces memory usage by 46% and 53% for MobileNetV2 and V3 IRBs, respectively. For MobileNetV3, we observe a 36% reduction in floating-point operations (FLOPs) when fine-tuning 5 blocks, while only incurring a 0.6% accuracy reduction on CIFAR10. Extensive experiments on multiple datasets demonstrate that our method is Pareto-optimal (best accuracy under given hardware constraints) compared to prior work in transfer learning for edge devices.