How to evaluate a trained ai like your replay? about elf HOT 5 CLOSED

You can turn on --save_prefix [your_prefix] to save the replay when you run run.py. Then, you can load them following instructions in README replay section.

from elf.

We have mcts baselines implemented, but it is currently broken after refactoring. We will fix it later.

from elf.

Thanks.
But after running run.py with --save_prefix [your_prefix], I got .bin file instead of .rep file.
Would you please give an instruction in detail? What is the command you use?

from elf.

Sorry, it appears that we removed that switch.
Uncomment game_MC/game.py line 69
# opt.save_replay_prefix = b"replay"

Then it will generate replay files for each game it runs. You might want to evaluate your model with a small number of games.

from elf.

Thank you. I got it.

from elf.