This is the official implementation of the Self Expanding Convolutional Neural Network paper, which is our submission to the ProjectX machine learning research competition organized by UofT AI
.
Project Structure
├─models
├── __init__.py
├── convBlock.py
├── dynamicCNN.py
├── identityConv.py
├── perceptron.py
├─utils
├── __init__.py
├── train.py
├── utils.py
├─ dataloaders.py
├─ main.py
├─ README.md
To Recreate the results:
- clone the current repository
- Install the following dependencies in a virtual environment:
- pytorch
- wandb
- tqdm
- Run the
main.py
file from the root directory
We benchmarked on the CIFAR 10 dataset. You can easily create your own custom dataset and pass the dataloaders into the train_model
function.
HAPPY CODING!!