use rllib v2.5.1 + optimise fetch raw data in gym environment

releat215 / releat

REinforcement LEarning for Algorithmic Trading is a python framework for medium frequency trading algorithms for MetaTrader 5

https://releat215.github.io/releat/

MIT License

22 stars 9 forks source link

Closed releat215 closed 9 months ago

releat215 commented 9 months ago

use rllib v2.5.1 because for v2.6 onwards there are architectural changes that affect logging, structure of agent_model.py and how rllib algorithms are used, i.e. impala. Updating to v2.6+ will require some syntax changes which will be done later
optimise how the raw data of each step is fetched from aerospike