Timezone: »

 
Poster
FracTrain: Fractionally Squeezing Bit Savings Both Temporally and Spatially for Efficient DNN Training
Yonggan Fu · Haoran You · Yang Zhao · Yue Wang · Chaojian Li · Kailash Gopalakrishnan · Zhangyang Wang · Yingyan Lin

Tue Dec 08 09:00 PM -- 11:00 PM (PST) @ Poster Session 2 #727
in Poster Session 3 »

Recent breakthroughs in deep neural networks (DNNs) have fueled a tremendous demand for intelligent edge devices featuring on-site learning, while the practical realization of such systems remains a challenge due to the limited resources available at the edge and the required massive training costs for state-of-the-art (SOTA) DNNs. As reducing precision is one of the most effective knobs for boosting training time/energy efficiency, there has been a growing interest in low-precision DNN training. In this paper, we explore from an orthogonal direction: how to fractionally squeeze out more training cost savings from the most redundant bit level, progressively along the training trajectory and dynamically per input. Specifically, we propose FracTrain that integrates (i) progressive fractional quantization which gradually increases the precision of activations, weights, and gradients that will not reach the precision of SOTA static quantized DNN training until the final training stage, and (ii) dynamic fractional quantization which assigns precisions to both the activations and gradients of each layer in an input-adaptive manner, for only "fractionally" updating layer parameters. Extensive simulations and ablation studies (six models, four datasets, and three training settings including standard, adaptation, and fine-tuning) validate the effectiveness of FracTrain in reducing computational cost and hardware-quantified energy/latency of DNN training while achieving a comparable or better (-0.12%~+1.87%) accuracy. For example, when training ResNet-74 on CIFAR-10, FracTrain achieves 77.6% and 53.5% computational cost and training latency savings, respectively, compared with the best SOTA baseline, while achieving a comparable (-0.07%) accuracy. Our codes are available at: https://github.com/RICE-EIC/FracTrain.

Author Information

Yonggan Fu (Rice University)
Haoran You (Rice University)
Yang Zhao (Rice University)
Yue Wang (Rice University)
Chaojian Li (Rice University)
Kailash Gopalakrishnan (IBM Research)
Zhangyang Wang (University of Texas at Austin)
Yingyan Lin (Rice University)

The assistant professor working on energy-efficient machine learning systems

More from the Same Authors