Parallel Deep Network CLI

A command-line interface to train various neural network architectures using asynchronous stochastic gradient descent (Hogwild!) using PyTorch.

Getting started

Create a virtual environment python3 -m venv env; source env/bin/activate and install dependencies pip install -r requirements.

You can then launch the application by running python src/main.py, which will train a feedforward neural network with 1 epoch, using 1 process and a batch size of 12. All logs will be saved in the src/logs folder.

Settings

To view available settings run python src/main.py --help.

Available options are:

--epochs INTEGER    number of epochs to train neural network.
--arch TEXT         neural network architecture to benchmark (conv or ff).
--distributed TEXT  whether to distribute data or not (y or n).
--procs INTEGER     number of processes to spawn.
--nodes INTEGER     number of cores to use.
--batches INTEGER   minibatch size to use.
--help              Show this message and exit.