brenjohn/brandubh

Using AlphaGo methods to play a game called Brandubh

brandubh's Issues

Update DualNet

The DualNet class should be updated to facilitate batched predictions

Memory leak while training with keras and tensorflow

There's a memory leak issue while training a zerobot in the parallel training script with keras 2.4.3 and tensorflow 2.4.1

The issue limits training as the training process is eventually killed by the OOM killer.

The master branch isn't protected

https://docs.github.com/en/repositories/configuring-branches-and-merges-in-your-repository/defining-the-mergeability-of-pull-requests/about-protected-branches

Add unit tests

It would be nice to have ample test coverage

rollout selection when choosing bots

The user should be able to select the number of evaluations to be made by the MCTS and Zero bots each turn as a way of setting the difficulty

Proper test coverage for the BrandubhView class

The BrandubhView class needs more tests.

This requires mocking the cursrs packages properly. See TODO comments in code.

Refactor brandubh UI to use model-view-controller design pattern

Refactor brandubh UI to use model, view and controller objects