mattdornfeld/moonraker

coinbase_ml

This is a monorepo containing code for training, evaluating, and serving a machine learning model for trading on the Coinbase Pro exchange.

Setup

Run

make install-dev

to install the project dependencies.

Building

To build the Scala portion of the project run

make build-scala

This will run the Scala test suite and build a fat jar containing the Scala portion of the project. The jar will be located in the top level of this repo. On the Python side only the proto files need to be built. To do so run

make build-python-protos

Testing

There are a few different tests you can run to verify the integrity of your changes (see the Makefile for a complete list). To run the Scala test suite run

make test-scala

To run the Python test suite run

make test-python

You can also run subsets of the Python test suite with the following commands

make test-unit test-mypy test-pylint

Running Locally

Last if you set the appropriate values for the environment variables DB_HOST, POSTGRES_USERNAME, and POSTGRES_PASSWORD you can run

python3 coinbase_ml/train/train.py -u

to execute an end to end test, using the test configs. It should take a minute or two to run.

Ideas This Repo is Based On

Continuous control with deep reinforcement learning

DeepLOB: Deep Convolutional Neural Networks for Limit Order Books

An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling

Human-level control through Deep Reinforcement Learning

Hierarchical Attention Networks for Document Classification

WaveNet: A Generative Model for Raw Audio