Installation

To enable importing of modules, from the parent directory execute:

pip install -e .

To install requirements: python -m pip install requirements.txt

To save requirements: python -m pip list --format=freeze --exclude-editable -f https://download.pytorch.org/whl/torch_stable.html > requirements.txt

• Note we use Python 3.9.4 for our experiments

The code in augerino_lib is an extension (with our modification) from the original Augerino code https://github.com/g-benton/learning-invariances. The code in data_augmentation/my_training.py is a modification of the pytorch example in https://github.com/pytorch/examples/tree/master/imagenet.

Running the code

For the experiments of Section 2:

To train the model update the appropriate result_dir (folder where to save the results) and data (your path to ImageNet data) in config_test_ddp_local.yaml and run python hydra_app_local.py. To test the model update the appropriate result_dir (folder where the trained models are) and job_id (id of the folder for a specific trained model you want to test) in config_test_local.yaml and run python hydra_test_local.py.

For remaining experiments:

Navigate to the corresponding directory, then execute: python run.py -m with the corresponding config.yaml file (which stores experiment configs).