notadamking/RLTrader

Making training/testing deterministic

Open

#36 opened on Jun 12, 2019

View on GitHub
 (0 comments) (0 reactions) (0 assignees)Python (1,592 stars) (535 forks)batch import
enhancementhelp wantedquestion

Description

Do you guys know how to make training / testing deterministic, e.g. train / test with reproducible results? There is a seed parameter in learn method, however, I don't think it works properly. Even if I set it to constant, the trained model comes out different every time. For testing, predict method has deterministic parameter, that seems to work as intended.

Contributor guide

Tech stack
python
Domain
machine learning
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
Python basicsReinforcement learning concepts
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
40
Research direction
Investigate the seed parameter in the learn method of RLTrader. Check if the seed is applied to all random number generators (numpy, random, TensorFlow/PyTorch). Look at the training loop and model initialization to ensure deterministic behavior. Refer to the training code in the repository for any missing seed setting.