Algorithmic crypto trading framework with Kafka and TensorFlow (Keras + TensorFlow Serving)
Documentation live at: https://carlomazzaferro.github.io/kryptoflow/index.html
STATUS: pre-alpha. Very active development. Feel free to rip the code apart, repurpose it, improve on it, or do whatever else it may please. Documentation is also not fully up to date. This will change soon.
The idea behind this project is to take care of the infrastructural work that goes behind building a machine learning-driven trading systems. In particular, the aim is giving users a working set up for:
- Gathering as granular data as possible
- Test a variety of strategies intuitively and with low friction
- Deploying models easily and keeping them up to date
- Trading continuously and reliably
All this done through a rich API and a browser-based GUI. Want to test your newly developed deep learning model tplace to start.
Another use case: want to build something like Quantopian but using cutting-edge machine learning? Then you may need to build it yourself. Here you'll find most of the tooling needed.
Some basic requirements:
- python3.6
- Docker
More at: https://carlomazzaferro.github.io/kryptoflow/installation.html
$ pip install kryptoflow$ kryptoflow init --name make-money --path /home/user/projectsThis will create all the configuration necessary to get going, alongside docker files and other stuff. Cd into the repo and check out its contents.
From the project directory, run:
docker-compose up$ kryptoflow train --epocs 10 --model model.pyThen
$ kryptoflow serve --model-number 1