Update default net to nn-735bba95dec0.nnue #3685
Conversation
|
@vondele The regression book is quite balanced. I suspect that gains in UHO book may not transfer well into the regression book. But let's see. Maybe a regression test after committing this? I'm 100% in favor of using UHO book though because this is what matters for TCEC. |
|
@SFisGOD I think so as well. BTW, could you update the description to include links to the used SPSA tests, as you did previously, that's kind of nice to have. |
Okay, I will update the PR. |
|
@vondele PR updated. The intermediate nets were not tested because I know that it's highly unlikely that these will pass as the number of games for the intermediate SPSAs were much lower than what I would usually give it. So no need to spend resources testing those. Only the nets after SPSA 3 were tested. |
SPSA 1: https://tests.stockfishchess.org/tests/view/61286d8b62d20cf82b5ad1bd
Parameters: A total of 256 net biases were tuned (hidden layer 2)
Base net: nn-33495fe25081.nnue
New net: nn-83e3cf2af92b.nnue
SPSA 2: https://tests.stockfishchess.org/tests/view/6129cf2162d20cf82b5ad25f
Parameters: A total of 64 net biases were tuned (hidden layer 1)
Base net: nn-83e3cf2af92b.nnue
New net: nn-69a528eaef35.nnue
SPSA 3: https://tests.stockfishchess.org/tests/view/612a0dcb62d20cf82b5ad2a0
Parameters: 256 net weights and 8 net biases (output layer)
Base net: nn-69a528eaef35.nnue
New net: nn-735bba95dec0.nnue
STC:
LLR: 2.94 (-2.94,2.94) <-0.50,2.50>
Total: 95144 W: 24310 L: 23999 D: 46835
Ptnml(0-2): 232, 11059, 24748, 11232, 301
https://tests.stockfishchess.org/tests/view/612bb3be0fdf40644b4b9996
LTC:
LLR: 2.94 (-2.94,2.94) <0.50,3.50>
Total: 33632 W: 8522 L: 8271 D: 16839
Ptnml(0-2): 18, 3511, 9516, 3744, 27
https://tests.stockfishchess.org/tests/view/612ce5b9bb4956d8b78eb5b3
Closes #3685
Bench: 5600615