DNN-Chip Predictor

This is the official implemenatation of DNN-Chip Predictor: An Analytical Performance Predictor for DNN Accelerators with Various Dataflows and Hardware Architectures [ICASSP'20]

Example to predict the enengy and latency given the operation on Eyeriss

Add your operation to the OPs_list in predictor.py

• "idx": the index of the operation in the operation set to be processed by predictor.py
• "type": the type of the operation, should be one from ["Conv", "AvgP", "MaxP", "FC"]
• "kernel_size": the kernel size of the operation, same with the kernel_size in PyTorch API
• "stride": the stride of the operation, same with the stride in PyTorch API
• "padding": the padding size of the operation, same with the padding in PyTorch API
• "input_H": the heigh of input feature map
• "input_W": the width of input feature map
• "input_C": the channel of input feature map
• "output_E": the heigh of output feature map
• "output_F": the width of output feature map
• "output_M": the channel of output feature map

run the predictor.py

python predictor.py

Publication

If you use this github repo, please cite:

@inproceedings{zhao2020dnn,
author = {Zhao, Yang and Li, Chaojian and Wang, Yue and Xu, Pengfei and Zhang, Yongan and Lin, Yingyan},
year = {2020},
month = {05},
pages = {1593-1597},
booktitle={International Conference on Acoustics, Speech, and Signal Processing},
title = {DNN-Chip Predictor: An Analytical Performance Predictor for DNN Accelerators with Various Dataflows and Hardware Architectures},
doi = {10.1109/ICASSP40776.2020.9053977}
}