Neptune cloud is the easiest way to start experimenting. Environment is already prepared, hence you care only about experiments.

1. clone this repo

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

2. register on the neptune.ml
3. install neptune-cli

$ pip3 install neptune-cli

4. log in to neptune cloud via command line

$ neptune login

5. 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.

6. run first experiment

$ neptune send experiment_manager.py --environment keras-2.0-gpu-py3 --worker gcp-gpu-medium --config neptune_config.yaml -- train_evaluate_predict_pipeline --pipeline_name glove_lstm

Now you can observe the progress of your experiment on Neptune dashboard. When its done, you can collect output from the Browse Files (left bar in the Neptune dashboard).

Check Neptune documentation for more options and our Wiki for detailed explanation of this starter code.

1. Entire dataset is already uploaded to the Neptune Cloud (/public directory), and ready to use.
2. Moreover the weights of the glove840 model are uploaded as well (/public directory). You are free to use this model in your solution.