Before running the code ...

downlad the data from https://drive.google.com/file/d/1YZQFSonulXuagMIfbJkZeTFJ6qEUuUxL/view?usp=sharing and extract files to a ./data folder

Do something like this .
├── data
│   ├── dataset.json
│   ├── imagesTr
│   ├── imagesTs
│   └── labelsTr
├── V_NAS.py
├── main.py
...
...
...

V_NAS.py

Model articture for VNAS, this model has ability to update specific parameters

searched_nas.py

Model articture for searched NAS, you can try differnet combination of model(s) by passing the desired block index list to each Encoder/Decoder

best.py & worst.py

Train the best or worst model

search.py

Differentialable articture search