Windows 64-bit WB, Windows 32-bit WB, and 64-bit UCI included (thanks to Odd Gunnar Malin's Wb2Uci adapter). If you want 32-bit UCI, edit Wb2Uci.eng, and replace all "w64" by "w32".
This version is 320 Elo stronger than the last in self-play..........
Roughly how strong do you believe that Giraffe is now (in CCRL terms).
Graham.
Hi Graham,
Depending on the time control, it seems to be about exactly as strong as Sungorus 1.4, so I would say about 2300. My guess is it's stronger at longer time control than shorter, since the engine is almost entirely based on eval. However, I haven't really had time to test it yet at longer time controls.
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.
Windows 64-bit WB, Windows 32-bit WB, and 64-bit UCI included (thanks to Odd Gunnar Malin's Wb2Uci adapter). If you want 32-bit UCI, edit Wb2Uci.eng, and replace all "w64" by "w32".
This version is 320 Elo stronger than the last in self-play..........
Roughly how strong do you believe that Giraffe is now (in CCRL terms).
Graham.
Hi Graham,
Depending on the time control, it seems to be about exactly as strong as Sungorus 1.4, so I would say about 2300. My guess is it's stronger at longer time control than shorter, since the engine is almost entirely based on eval. However, I haven't really had time to test it yet at longer time controls.
Thanks. Great progress!
Can I use a generic opening book with Giraffe?
Windows 64-bit WB, Windows 32-bit WB, and 64-bit UCI included (thanks to Odd Gunnar Malin's Wb2Uci adapter). If you want 32-bit UCI, edit Wb2Uci.eng, and replace all "w64" by "w32".
This version is 320 Elo stronger than the last in self-play..........
Roughly how strong do you believe that Giraffe is now (in CCRL terms).
Graham.
Hi Graham,
Depending on the time control, it seems to be about exactly as strong as Sungorus 1.4, so I would say about 2300. My guess is it's stronger at longer time control than shorter, since the engine is almost entirely based on eval. However, I haven't really had time to test it yet at longer time controls.
Thanks. Great progress!
Can I use a generic opening book with Giraffe?
Yeah sure!
It seems to play openings quite well, but using an opening book will save some thinking time.
I'll probably have it generate its own opening book for the next release.
Oh yeah and by the way, the scores it report are probabilistic and not based on centipawns. +100.00 means it thinks it will win for sure, and -100.00 means it thinks it will lose for sure. So if your GUI does any adjudication based on engine-reported scores, strange things may happen. Let me know if that's the case. I can make a special build with scaled score reporting or something.
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.
Windows 64-bit WB, Windows 32-bit WB, and 64-bit UCI included (thanks to Odd Gunnar Malin's Wb2Uci adapter). If you want 32-bit UCI, edit Wb2Uci.eng, and replace all "w64" by "w32".
This version is 320 Elo stronger than the last in self-play..........
Roughly how strong do you believe that Giraffe is now (in CCRL terms).
Graham.
Hi Graham,
Depending on the time control, it seems to be about exactly as strong as Sungorus 1.4, so I would say about 2300. My guess is it's stronger at longer time control than shorter, since the engine is almost entirely based on eval. However, I haven't really had time to test it yet at longer time controls.
Thanks. Great progress!
Can I use a generic opening book with Giraffe?
Yeah sure!
It seems to play openings quite well, but using an opening book will save some thinking time.
I'll probably have it generate its own opening book for the next release.
Oh yeah and by the way, the scores it report are probabilistic and not based on centipawns. +100.00 means it thinks it will win for sure, and -100.00 means it thinks it will lose for sure. So if your GUI does any adjudication based on engine-reported scores, strange things may happen. Let me know if that's the case. I can make a special build with scaled score reporting or something.
Please make sure that own book can be tuned off in the options.
Can it run on non-popcount computers as well as the popcount ones?
Windows 64-bit WB, Windows 32-bit WB, and 64-bit UCI included (thanks to Odd Gunnar Malin's Wb2Uci adapter). If you want 32-bit UCI, edit Wb2Uci.eng, and replace all "w64" by "w32".
This version is 320 Elo stronger than the last in self-play..........
Roughly how strong do you believe that Giraffe is now (in CCRL terms).
Graham.
Hi Graham,
Depending on the time control, it seems to be about exactly as strong as Sungorus 1.4, so I would say about 2300. My guess is it's stronger at longer time control than shorter, since the engine is almost entirely based on eval. However, I haven't really had time to test it yet at longer time controls.
Thanks. Great progress!
Can I use a generic opening book with Giraffe?
Yeah sure!
It seems to play openings quite well, but using an opening book will save some thinking time.
I'll probably have it generate its own opening book for the next release.
Oh yeah and by the way, the scores it report are probabilistic and not based on centipawns. +100.00 means it thinks it will win for sure, and -100.00 means it thinks it will lose for sure. So if your GUI does any adjudication based on engine-reported scores, strange things may happen. Let me know if that's the case. I can make a special build with scaled score reporting or something.
Please make sure that own book can be tuned off in the options.
Can it run on non-popcount computers as well as the popcount ones?
The UCI binary is based on Wb2Uci, so that should be fine.
It does work without popcount.
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.
Windows 64-bit WB, Windows 32-bit WB, and 64-bit UCI included (thanks to Odd Gunnar Malin's Wb2Uci adapter). If you want 32-bit UCI, edit Wb2Uci.eng, and replace all "w64" by "w32".
This version is 320 Elo stronger than the last in self-play..........
Roughly how strong do you believe that Giraffe is now (in CCRL terms).
Graham.
Hi Graham,
Depending on the time control, it seems to be about exactly as strong as Sungorus 1.4, so I would say about 2300. My guess is it's stronger at longer time control than shorter, since the engine is almost entirely based on eval. However, I haven't really had time to test it yet at longer time controls.
Thanks. Great progress!
Can I use a generic opening book with Giraffe?
Yeah sure!
It seems to play openings quite well, but using an opening book will save some thinking time.
I'll probably have it generate its own opening book for the next release.
Oh yeah and by the way, the scores it report are probabilistic and not based on centipawns. +100.00 means it thinks it will win for sure, and -100.00 means it thinks it will lose for sure. So if your GUI does any adjudication based on engine-reported scores, strange things may happen. Let me know if that's the case. I can make a special build with scaled score reporting or something.
Please make sure that own book can be tuned off in the options.
Can it run on non-popcount computers as well as the popcount ones?
The UCI binary is based on Wb2Uci, so that should be fine.
It does work without popcount.
Will include it in my next appropriate tournament.
Graham Banks wrote:Will include it in my next appropriate tournament.
Awesome. Thanks a bunch!
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.
That's not configurable, yet. It's hardcoded at 256MB right now.
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.
