alexx-ftw

Results 20 comments of alexx-ftw

Hello I am using this library and it is really good. Thanks for the hard work. I trained 6 DQN models on NLH with 1_000_000 episodes, evaluating every 10_000 with...

@daochenzha Thanks for the quick reply. I came to the same "eval_step" function by going backwards from the training example into the functions and classes. I printed the state variable...

> If you would like to improve NLH. You may need to do feature engineering by modifying https://github.com/datamllab/rlcard/blob/master/rlcard/envs/nolimitholdem.py Also, you may try training with multiple GPUs using [DMC](https://github.com/datamllab/rlcard/blob/master/examples/run_dmc.py). > >...

> By the way, how is the "obs" key structured? Looks like it would be the cards on the deck but I am not sure as it has 54 instead...

> @alexx-ftw did you ever get this working? how well does your model play? I am running a social poker plattform and I'm looking for an AI that I can...

> I don't quite understand, I thought with the help of this project you could train a network that basically would take a table-state as input and give you a...

I would love Firefox support. Would be a pretty good differentiation factor

There are many wallets other than Rabby. I refuse to use Brave or any other crappy Browser