Getting started: local instance

Prerequisites

1. Python 3.5
2. Neptune

Installation

clone this repo to your working directory

$ git clone https://github.com/neptune-ml/kaggle-toxic-starter.git

create virtualenv with python 3.5

$ virtualenv toxic_venv -p python3.5

activate newly created virtualenv

$ source toxic_venv/bin/activate

install TensorFlow and Keras.

• in case you have GPU

$ pip3 install tensorflow-gpu==1.1.0
$ pip3 install Keras==2.0.8

• in case you do not have GPU

$ pip3 install tensorflow==1.1.0
$ pip3 install Keras==2.0.8

install remaining requirements (note that neptune is included in the requirements)

$ pip3 install -r requirements.txt

Neptune setup

• register on the neptune site to receive $5 in GPU time.
• create neptune project
  • go to the neptune site and log in
  • create new project named toxic. To do it follow the link Projects (top bar, left side), then click New project button. This action will generate project-key TOX, which is already listed in the neptune_config.yaml.

log in to neptune via command line

$ neptune login

Create train and validation split

• define paths in the neptune.yaml file

  data_dir: path/to/my/data/dir

• run train test splitting script

neptune run -- train_valid_split

At this point you are ready to go! :-)