Needs Simulator so we don't need the actual game to train data.

peter1591 / hearthstone-ai

A Hearthstone AI based on Monte Carlo tree search and neural nets written in modern C++.

302 stars 49 forks source link

Needs Simulator so we don't need the actual game to train data. #94

Closed SethKitchen closed 6 years ago

SethKitchen commented 6 years ago

I can run this on a cluster with many GPU to train if you implement simulator in c++ or c#

SethKitchen commented 6 years ago

https://github.com/HearthSim/SabberStone

SethKitchen commented 6 years ago

I will attempt to integrate

SethKitchen commented 6 years ago

Oh I think your judgement framework can already simulate? But I want to use random decks until it learns what is the best deck too. Can you make wiki on how to train it in this way?