Werewolf wrote:
When I load this in the Fritz GUI it just says "empty" in the engine name pane and won't load.
I have just had correspondence with ChessBase and they told me that 64-bit Fritz versions do not support 32-bit engines. I had the same 'empty' fields in F15 when there was no problem under Fritz 13.
Its a pity there is not a 64bit version of Wb2Uci.exe available.
That would presumably enable 32bit (and 64bit) WB engines to work in a Fritz 64bit GUI.
32bit UCI engines could then use polyglot.exe in conjunction with the 64bit Wb2Uci.exe
I use Wb2Uci.exe to enable WB engines such as Crafty 64bit to work in my Fritz 13 32bit GUI using a 64bit OS.
Werewolf wrote:Sorry, please ignore the last message, it was a GUI error.
However, is there a new Giraffe on the way?
Technically maybe. But I have so little time to work on Giraffe now it's probably not going to happen any time soon.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.
Werewolf wrote:Sorry, please ignore the last message, it was a GUI error.
However, is there a new Giraffe on the way?
Technically maybe. But I have so little time to work on Giraffe now it's probably not going to happen any time soon.
This shows I don't understand NNs. I was under the impression you could just let the thing train by itself and it would get better and better. I guess it's not so simple...
Werewolf wrote:Sorry, please ignore the last message, it was a GUI error.
However, is there a new Giraffe on the way?
Technically maybe. But I have so little time to work on Giraffe now it's probably not going to happen any time soon.
This shows I don't understand NNs. I was under the impression you could just let the thing train by itself and it would get better and better. I guess it's not so simple...
Ah no it's much more complicated than that. Each released version has been trained until convergence, so further training won't help. Improvements have to come from the algorithm, training procedure, data, or better neural network architecture.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.