NEW RELEASE: we released our new, engineered and user-friendly DNAS library named PLiNIO which includes channel-wise precision assignement among the different implemented methods. We highly suggest to try this new release for your experiments!

Copyright (C) 2022 Politecnico di Torino, Italy. SPDX-License-Identifier: Apache-2.0. See LICENSE file for details.

Authors: Matteo Risso, Alessio Burrello, Luca Benini, Enrico Macii, Massimo Poncino, Daniele Jahier Pagliari

multi-prec-nas

Reference

If you use our code in your experiments, please make sure to cite our paper:

@INPROCEEDINGS{9969373,
  author={Risso, Matteo and Burrello, Alessio and Benini, Luca and Macii, Enrico and Poncino, Massimo and Pagliari, Daniele Jahier},
  booktitle={2022 IEEE 13th International Green and Sustainable Computing Conference (IGSC)}, 
  title={Channel-wise Mixed-precision Assignment for DNN Inference on Constrained Edge Nodes}, 
  year={2022},
  volume={},
  number={},
  pages={1-6},
  doi={10.1109/IGSC55832.2022.9969373}}

Datasets

The current version support the following datasets and tasks taken from the benchmark suite MLPerf Tiny:

• CIFAR10 - Image Classification.
• MSCOCO - Visual Wake Words.
• Google Speech Commands v2 - Keyword Spotting.
• ToyADMOS - Anomaly Detection

How to run

Image Classification

1. Visit the folder: cd image_classification.
2. Run the provided shell script run_ic.sh:

source run_ic.sh <regularization_strenght> 0 resnet8_w248a248_multiprec search ft

Visual Wake Words

1. Run the provided Makefile to download the desired dataset: make vww-init.
2. Visit the folder: cd visual_wake_words.
3. Run the provided shell script run_vww.sh:

source run_vww.sh <regularization_strenght> 0 mobilenetv1_w248a248_multiprec search ft

Keyword Spotting

1. Visit the folder: cd keyword_spotting.
2. Run the provided shell script run_kws.sh:

source run_kws.sh <regularization_strenght> 0 dscnn_w248a248_multiprec search ft

Anomaly Detection

1. Run the provided Makefile to download the desired dataset: make andet-init.
2. Visit the folder: cd anomaly_detection.
3. Run the provided shell script run_andet.sh:

source run_andet.sh <regularization_strenght> 0 denseae_w248a248_multiprec search ft

License

This code is released under Apache 2.0, see the LICENSE file in the root of this repository for details.