Decrease LMR if the move gives check

The larger 24x64 net architecture is failing STC so we instead try 16x96. This has far fewer parameters but still has a 5% slowdown we need to overcome. Based on training metrics this lies squarely between the baseline and 24x64.

SIMD movepicker + clamp scores to int16 to compress

Having passed the fixed node test as expected, we try an STC test with [-5, 0] bounds, as the slowdown is expected to hurt more than in actual play.

Larger net, hopefully a clear improvement at fixed nodes. Around 16% less Nps.

3.6.12: fix tb tt bug (#324)

Decrease LMR if the move gives check

Prefetch TT entry before apply_move using precomputed key AVX2: +2.10% +/- 0.19%

Update default network to ed164cf2.nn (#890)

768 l1

768 l1

768 l1

Trying to better understand the hybrid loss. Both nets are not fully trained, but are at a similar point in training.

Trying to better understand the hybrid loss. Both nets are not fully trained, but are at a similar point in training. 311n removes the CE loss component completely. Even if worth Elo, we may not want this.

Final net with less CE loss component.

Reduced the weighting of the draw prediction loss component.

Increase Negative Singular Extensions by one ply for PV nodes Bench: 4,382,866

check if non-Syzygy data does better with access to Syzygy than without

9.3B new data

Draw calibration handled with MSE loss directly on the draw probability should be less sensitive to outliers that are common in the high entropy STC time controls the training data stems from.

Fully trained model with identical hyperparameters to prior gen, but extra data. This should clearly beat the master branch.

3.6.10: fix cache entry

3.6.10: fix cache entry

3.6.10: fix cache entry

Testing net with same hyperparameters but extra self play data a bit prematurely to get a sense of where things lie.