Skip to content

An OpenAI gym using crypto currency trading data as its observational space with actions: hodl, market_buy, market_sell, limit_buy, limit_sell, amount, and quantity.

License

Notifications You must be signed in to change notification settings

mellertson/crypto-trading-gym

Repository files navigation

Crypto Trading Open AI Gym

CryptoTradingGym is a collection of OpenAI Gym environments for reinforcement learning-based trading algorithms. It was forked from AnyTrading.

CryptoTradingGym is intended to be used with another project I authored, namely nupic_predictor.

Required Dependencies

This project is dependant on the following external projects:

  • nupic-predictor
  • Crypto Ninja REST API (api)

Running the Crypto Gym on LOCALHOST

  1. Install required pip packages using python3 -m pip install numpy cython pyparsing==2.4.7
  2. Install the Crypto Gym by executing python ./setup.py install
  3. Run the test case in Test_QLearnAgent_class.test_run_the_agent_for_one_episode

Building The Docker Container

Because, Docker Compose is used you can build the container by executing the following in a terminal.

docker-compose build

Deploying on JONIN (production)

To begin running the Nupic Q-Learning Trading Agent executing the following on a Docker Swarm manager node.

docker stack deploy -c stack.jonin.yaml --with-registry-auth bamm-crypto-trading-gym

Once launched the you can view the agents progress training in the Crypto Trading Gym at https://testnet.bitmex.com.

About

An OpenAI gym using crypto currency trading data as its observational space with actions: hodl, market_buy, market_sell, limit_buy, limit_sell, amount, and quantity.

Resources

License

Stars

Watchers

Forks

Packages

No packages published