Aughie Python Client Library

The Aughie Python client library provides access to the Neural Network Database API.

The project is currently in a proof-of-concept phase.

Install

Install the library from the Python Package Index with pip:

pip install aughie

Aughie needs Python 3.5 or later.

Example

This code snippet returns all networks from the database that have a Conv2D layer with a tanh activation and a MaxPooling2D layer and presents the data as a Pandas data frame.

from aughie import nndb

client = nndb.ApiClient()
networks = client.get_networks(layers=[{'type': 'Conv2D', 'activation': 'tanh'},
                                       {'type': 'MaxPooling2D'}])
df = networks.to_data_frame()

Tutorial

The Neural Network Database API Intro tutorial is a Jupyter notebook that shows how to use the database to analyze the relationship between optimizers and loss functions and introduces more fine-grained queries such as searching for networks used for image generation based on layer patterns.

You can also run the tutorial online directly from your browser by following the link to Google Colab.

Docs

The documentation is available at: https://aughie-py.readthedocs.io