search

happypepper / DeepHoldem

203 stars 65 forks source link

Samples count #1

Open snarb opened 6 years ago

snarb commented 6 years ago

Hi! How much samples did you use to train the network? What was the huber loss? Can You share the training data to play with it and compare with my? (I have implimented deepStack in c++ now generating samples)

happypepper commented 6 years ago

I've added the training details to the readme. You can also observe the validation losses yourself by going to Data/Models/NoLimit/<street> and th, then torch.load('final_cpu.info') in the interactive terminal.

snarb commented 6 years ago

Thanks. Didn't you measure the loss after the buckets CFR ->hole hands CFR conversion, not for the buckets? Bucketing by itself should add extra loss. It is interesting how much.

happypepper commented 6 years ago

Actually, the loss was measured after the hole -> buckets conversion, so I think the real loss will be higher than the reported loss (but I'm not sure). I agree it would be interesting to measure the difference.