google-deepmind / hanabi-learning-environment

hanabi_learning_environment is a research platform for Hanabi experiments.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[Feature Request] Start game from specific state

hellovertex opened this issue · comments

Hi!

We are a group of researchers in TU Berlin, and we are currently working on MARL using Hanabi as a testbed. There are some ideas that we would like to try, but this will require to be able to initialize the environment from a specific state that we choose and not from a new game state. Ideally that would include arranging the cards on the deck as well. For a subset of the things that we would like to try, just being able to save and load a state will be enough. Do you consider adding one of these features, so that we can use tree search methods?

Hi,

That sounds like a great addition. There aren't any immediate plans to do so, but let us know if you'd like to take ownership of this.