APQ: Joint Search for Network Architecture, Pruning and Quantization Policy

TL;DR

APQ introduces a joint search method for neural architecture, pruning, and quantization, utilizing a transfer learning approach for accuracy prediction to enable efficient, environmentally friendly model optimization on resource-limited hardware.

Contribution

It proposes a novel joint optimization framework for architecture, pruning, and quantization, with a transfer learning-based accuracy predictor to improve efficiency and reduce training costs.

Findings

Achieves 2x latency reduction on ImageNet with same accuracy.

Reduces GPU hours and CO2 emissions compared to separate optimization.

Outperforms existing methods in accuracy and efficiency.

Abstract

We present APQ for efficient deep learning inference on resource-constrained hardware. Unlike previous methods that separately search the neural architecture, pruning policy, and quantization policy, we optimize them in a joint manner. To deal with the larger design space it brings, a promising approach is to train a quantization-aware accuracy predictor to quickly get the accuracy of the quantized model and feed it to the search engine to select the best fit. However, training this quantization-aware accuracy predictor requires collecting a large number of quantized <model, accuracy> pairs, which involves quantization-aware finetuning and thus is highly time-consuming. To tackle this challenge, we propose to transfer the knowledge from a full-precision (i.e., fp32) accuracy predictor to the quantization-aware (i.e., int8) accuracy predictor, which greatly improves the sample efficiency. Besides, collecting the dataset for the fp32 accuracy predictor only requires to evaluate neural networks without any training cost by sampling from a pretrained once-for-all network, which is highly efficient. Extensive experiments on ImageNet demonstrate the benefits of our joint optimization approach. With the same accuracy, APQ reduces the latency/energy by 2x/1.3x over MobileNetV2+HAQ. Compared to the separate optimization approach (ProxylessNAS+AMC+HAQ), APQ achieves 2.3% higher ImageNet accuracy while reducing orders of magnitude GPU hours and CO2 emission, pushing the frontier for green AI that is environmental-friendly. The code and video are publicly available.