mindspore/model_zoo/research/cv/HourNAS

Contents

• HourNAS Description
• Model Architecture
• Dataset
• Environment Requirements
• Script Description
  • Script and Sample Code
    • Training Process
    • Evaluation Process
• Model Description
  • Performance
    • Evaluation Performance
• Description of Random Situation
• ModelZoo Homepage

HourNAS Description

HourNAS is an efficient neural architecture search method. Only using 3 hours (0.1 days) with one GPU, HourNAS can search an architecture that achieves a 77.0% Top-1 accuracy, which outperforms the state-of-the-art methods.

Paper: Zhaohui Yang, Yunhe Wang, Xinghao Chen, Jianyuan Guo, Wei Zhang, Chao Xu, Chunjing Xu, Dacheng Tao, Chang Xu. HourNAS: Extremely Fast Neural Architecture Search Through an Hourglass Lens. In CVPR 2021.

Model architecture

The overall network architecture of HourNAS is show below:

Link

Dataset

Dataset used: CIFAR-10

• Dataset size：175M，60,000 32*32 colorful images in 10 classes
  • Train：146M，50,000 images
  • Test：29M，10,000 images
• Data format：binary files
  • Note：Data will be processed in src/dataset.py

Environment Requirements

• Hardware (GPU)
• Framework
  • MindSpore
• For more information, please check the resources below:
  • MindSpore Tutorials
  • MindSpore Python API

Script Description

Script and Sample Code

.HourNAS
├── README.md               # descriptions about HourNAS
├── src
│   ├── architectures.py    # definition of HourNAS-F model
│   ├── dataset.py          # data preprocessing
│   ├── hournasnet.py       # HourNAS general architecture
│   └── utils.py            # utility functions
├── eval.py                 # evaluation interface

Training process

To Be Done

Evaluation Process

Launch

# infer example

python eval.py --model hournas_f_c10 --dataset_path [DATA_PATH] --GPU --ckpt [CHECKPOINT_PATH]

Result

result: {'Top1-Acc': 0.9618389423076923} ckpt= ./hournas_f_cifar10.ckpt

Model Description

Performance

Evaluation Performance

Model	FLOPs (M)	Params (M)	ImageNet Top-1
MnasNet-A1	312	3.9	75.2%
HourNAS-E	313	3.8	75.7%
EfficientNet-B0	390	5.3	76.8%
HourNAS-F	383	5.3	77.0%

More details in Paper.

Description of Random Situation

We set the seed inside dataset.py.

ModelZoo Homepage

Please check the official homepage.