Examples

• PCAX comes with examples that showcase how to implement different architectures using the library.
• Examples also include hypertuning code, search spaces, and the best parameters, which makes it easy to experiment and reproduce the results from the paper.
• Both PC and BP versions of architectures are provided for performance and implementation comparison.

Common setup

Python dependencies for all examples are specified in the examples group in the pyproject.toml file. This dependency group is installed automatically when you run poetry install. However, if you installed dependencies without the examples group, make sure to install it:

1. Activate the virtual environment: conda activate pcax.
2. cd in the root directory of pcax.
3. poetry install --no-root --with examples

If you want to run hypertuning, you will also need to install stune:

1. cd outside of the pcax root directory.
2. Clone the stune repository: git clone [email protected]:liukidar/stune.git.
3. cd stune
4. Activate the virtual environment: conda activate pcax.
5. pip install -e .
6. Install additional stune dependencies: pip install omegaconf-argparse names_generator redis optuna-dashboard
7. cd into the pcax/examples folder.
8. In a new terminal, launch Redis. For this, you need to install podman or docker: ./run_redis.sh. Return to the original terminal after this step.
9. You should do this step for each example subfolder!. Configure stune in pcax/examples/autoencoders/deconv folder: cd pcax/examples/autoencoders/deconv and python -m stune.config. Use 0.0.0.0:6379 for STUNE_HOST and empty for the rest.
10. In a new terminal, activate the pcax environment and run the optuna dashboard from the pcax/examples folder: ./run_dashboard.sh. The dashboard is available at http://127.0.0.1:8080.