HAPPYNET (HN)

Requirement

• PyTorch
• torchvision
• tabulate

Usage

To run HN use the following command:

python3 testac.py --dir=<DIR> --dataset=CIFAR100 --data_path=<PATH>  --model=PreResNet164 --epochs=225 --lr_init=0.1 --wd=3e-4 --swa --swa_start=126 --swa_lr=0.05

Parameters:

• DIR — path to training directory where checkpoints will be stored
• PATH — path to the data directory

Results

• Test accuracy (%) of HN on CIFAR-100 on PreResNet164 (150) is 80.4
• Our model has 1.7M parameters.

Train Model From Scratch

To train HN use the following command:

python3 train.py --dir=<DIR> --dataset=CIFAR100 --data_path=<PATH>  --model=PreResNet164 --epochs=225 --lr_init=0.1 --wd=3e-4 --swa --swa_start=126 --swa_lr=0.05

Our approach

We use the SWA method to train our model.