This repository contains the code for the paper:
"Neural Growth and Pruning in Dynamic Learning Environments." Kaitlin Maile, Hervé Luga, and Dennis G. Wilson. Second Conference on Automated Machine Learning (Workshop Track). 2023.
Make and activate a clean conda or virtual envirnoment. Clone NeurOps and add it to your PYTHONPATH. Then, run the following command from the root directory of this repository:
python3 -m pip install -r requirements.txt
The following datasets are used in this repository. Follow each link to documentation on how to download the dataset. Place the datasets in a data folder, which by default is assumed to be in the same directory as this repository but can be specified elsewise using the --path argument.
The following pretrained models are used in this repository. Each link will directly download the model weights. Place these files in the same data folder as the datasets.
To run an experiment, use the following command:
python3 src/transfer.py
See the file src/transfer.py for the full list of arguments.