Installation

This page covers installation steps for the open solution:

1. Get the code
2. Create a Python virtualenv
3. Install Pytorch
4. Install the rest of the open solution requirements
5. Register on Neptune (optional)

Prerequisites

• Python 3.5

Note: This open solution is developed and tested on Ubuntu 16.04 machines only. If you run into unexpected issues while running the solution on another OS - consider adding a GitHub issue on our repo.

1. Get the code

Clone or download DSB open solution code. Type:

$ cd your/working/directory
$ git clone https://github.com/minerva-ml/open-solution-data-science-bowl-2018

Note that you must have writing permission in your working directory.

2. Create a Python virtualenv

2.1. Install virtualenv

Virtualenv is a tool we recommend for creating isolated Python environments. It can be installed as follows:

$ [sudo] pip3 install virtualenv

2.2 Create a virtualenv for the DSB 2018 open solution:

$ cd your/working/directory
$ virtualenv dsb_open_solution -p python3.5

(Note that you must have writing permission in your working directory.)

2.3. Activate the newly created environment:

$ source dsb_open_solution/bin/activate

3. Install PyTorch

The DSB open solution uses PyTorch, which comes in two flavours: with and without GPU support. Due to computational burden of deep learning models used in this solution, it is highly recommended to train your solution on a GPU.

If you don't have a GPU on your local machine, you can use cloud GPU instances (e.g. through Neptune, using your free $5 credit).

Find the correct way of installing PyTorch on your system by following PyTorch Get Started. Make sure that you picked Python 3.5 and the appropriate CUDA version (or no CUDA if you don't have a local GPU).

4. Install the rest of open solution requirements

Once your environment is activated (see Setup virtualenv section), install the remaining requirements:

$ pip3 install -r requirements.txt

5. Register on Neptune (optional)

Neptune is our experiments monitoring system. It makes it easy to track the progress of your experiments, check intermediate results, and reproduce results of every experiment you run.

To register on Neptune, follow these steps:

5.1. Go to https://neptune.ml/ and create your account.

5.2. Go to Neptune Dashboard and create a project called data_science_bowl with key DSB.

5.3. Log in to your Neptune account:

$ neptune account login

You get free $5 credit for your experiments.

At this point you are all set. Go to experimentation guideline to learn how to run your first experiment!