David J Wu

Hey there, I'm not sure what you want to do, but there are various options depending on your intent: * If you only want to contribute your computing power to...

Thanks for the question. 40b is probably not saturated, it probably will continue to improve at the same overall trend that it has been for a little while more. One...

Thanks for the tests! It sounds like we're close enough to time parity that we I agree with switching soon. I will get the release out supporting TensorRT, and then...

The idea of training KataGo networks on human games to get a set of nets that accurately predicted human play is something that I had planned myself to experiment with...

@acristescu https://github.com/lightvector/KataGo/blob/master/cpp/neuralnet/desc.cpp#L1111 Take a look at the code starting at that line and below. KataGo uses the extension to figure out whether it should try to load the file as...

Did you try adding the trtDeviceToUseThread0, etc. in the selfplay cfgs?

Got around to taking a look at this - neat idea, yeah. One question - since this change is fundamentally about not taking too much more time than the opponent,...

Oh, nice catch thanks! https://github.com/lightvector/KataGo/blob/master/cpp/search/search.cpp#L461 That this variable is declared as "double_t" instead of "double" is definitely a typo, it should just be "double", I don't think it was ever...

Thanks for the bump - heh, size_t being 4 bytes is interesting. I think there's some chance everything works if you simply remove or suppress that assert. Does that seems...

Thanks for working through this. I know that different system libraries may return slightly different answers for transcendental functions, but of course numerical analysis is hard so I just set...