DeepCFR on Leduc Question

[rpSebastian]

Hi,

I am running DeepCFR_Leduc experiment. I use the same hyperparameters recommended by the configuration file deep_cfr_leduc.cfg. When the number of iterations reaches 100, the exploitability reaches 0.31, which is higher than the corresponding value 0.1 in the paper.

I find that the hyperparameters of kuhn poker and leduc poker are the same. The performance on kuhn poker roughly matches, but I am not able to replicate the performance on leduc poker. Are you using different hyperparameters on the leduc poker?

Do you have any idea why this would happen? It would be nice if you could provide the source code including hyperparameters and random seeds to reproduce the same result.

Thank you very much.