A library built on top of TensorFlow, Sonnet, and Sacred to faciliate deep learning research.

Version: 0.1.0.dev1

• Python 3.6
• virtualenv is strongly recommended
• CUDA 9.0 and cuDNN (if enabling GPU)

Clone to repository, cd into the root directory, activate a virtual environment (optional), and run

pip install -r requirements.txt
pip uninstall tensorflow && pip install tensorflow-gpu # run this line if you want to enable GPU

Setup s more complicated than this on Crane, but talk to Ellie directly about it because we don't have an automated process nailed down yet.

Documentation on master can be found at sounds-deep.readthedocs.io or can be built by running

./docs/build_scrip.sh

and pointing a browser at ./docs_build/index.html

Because the use of this package is expected to stay within the lab right now, you can find me in person or on slack with any questions.

import sounds_deep as sd

• contrib.data: Easy downloading of standard datasets and loading for TensorFlow
• contrib.distributions: Handles distributions in ways not done in tf.contrib.distributions (use tfd when possible)
• contrib.experiments: Executable files with command line interfaces which train a model
• contrib.models: Sonnet modules implementing entire model frameworks
• contrib.parameterized_distributions: Distributions with parameters baked in
• contrib.sacred_ingredients: Classes inheriting from Sacred.Ingredient

Everyone in the lab is invited to contribute code pertaining to Sonnet, TensorFlow, or deep learning/machine learning with Python.

Eleanor Quint, a Ph.D. student in the computer science and engineering department at the University of Nebraska-Lincoln