BitSplit¶
Abstract¶
Network quantization is essential for deploying deep models to IoT devices due to its high efficiency. Most existing quantization approaches rely on the full training datasets and the time-consuming fine-tuning to retain accuracy. Posttraining quantization does not have these problems, however, it has mainly been shown effective for 8-bit quantization due to the simple optimization strategy. In this paper, we propose a Bit-Split and Stitching framework (Bit-split) for lower-bit post-training quantization with minimal accuracy degradation. The proposed framework is validated on a variety of computer vision tasks, including image classification, object detection, instance segmentation, with various network architectures. Specifically, Bit-split can achieve near-original model performance even when quantizing FP32 models to INT3 without fine-tuning.
Results and models¶
Model |
Params(M) |
Flops(G) |
Top-1 (%) |
Top-5 (%) |
Config |
Download |
|---|---|---|---|---|---|---|
## Citation |
@inproceedings{10.5555/3524938.3525851,
author = {Wang, Peisong and Chen, Qiang and He, Xiangyu and Cheng, Jian},
title = {Towards Accurate Post-Training Network Quantization via Bit-Split and Stitching},
year = {2020},
publisher = {JMLR.org},
abstract = {Network quantization is essential for deploying deep models to IoT devices due to its high efficiency. Most existing quantization approaches rely on the full training datasets and the time-consuming fine-tuning to retain accuracy. Posttraining quantization does not have these problems, however, it has mainly been shown effective for 8-bit quantization due to the simple optimization strategy. In this paper, we propose a Bit-Split and Stitching framework (Bit-split) for lower-bit post-training quantization with minimal accuracy degradation. The proposed framework is validated on a variety of computer vision tasks, including image classification, object detection, instance segmentation, with various network architectures. Specifically, Bit-split can achieve near-original model performance even when quantizing FP32 models to INT3 without fine-tuning.},
booktitle = {Proceedings of the 37th International Conference on Machine Learning},
articleno = {913},
numpages = {10},
series = {ICML'20}
}