Training Related Questions


we joined the competition a little late. We screened the questions as much as we can. I hope we dont ask the same questions again :slight_smile: Here are a few questions from our side:

Do we have training slot after the submission to learn the different environments or shall we train the model prior to submission?

As we read there will be 1000 environments for evaluation. Will all they have a different map? Or are the maps being reused (resp. will there be only one map used in all environments)?

Can the controller code access cloud such as a model residing in Google Cloud Platform or should the whole model reside in the AI crowd server?

Thanks in advance!

Hi @leventkent

Welcome to the Challenge :steam_locomotive::railway_car::railway_car::railway_car::railway_car::confetti_ball::tada:

Here the answers to your questions:

  • Given the limited time of each submission we recommend that you submit already trained models.
  • In the current round there are actually only 250 envs that you have to solve. They all have different maps and different schedules for individual agents.
  • They have to reside within the repository.

Hope this helps you get started with the challenge, have fun.

The Flatland Team

Additional information:

You can use git-lfs to store large model files