Dev Builds » 20230531-0651

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 7. This yields an approximate Elo difference and establishes confidence in the strength of the dev builds.

Summary

Host Duration Avg Base NPS Games Wins Losses Draws Elo
ncm-et-3 09:27:23 1954302 3338 2858 3 477 +443.18 +/- 15.59
ncm-et-4 09:26:40 1954308 3315 2892 3 420 +465.3 +/- 16.63
ncm-et-9 09:27:19 1953434 3332 2894 8 430 +457.73 +/- 16.45
ncm-et-10 09:27:05 1953302 3318 2837 2 479 +442.06 +/- 15.55
ncm-et-13 09:27:30 1959237 3359 2871 10 478 +438.62 +/- 15.6
ncm-et-15 09:27:07 1951078 3338 2887 5 446 +453.93 +/- 16.14
20000 17239 31 2730 +449.89 +/- 6.51

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
190142 ncm-et-4 2023-06-01 14:21 00:53:25 1951628 315 280 0 35 +492.17 +/- 59.33
190141 ncm-et-10 2023-06-01 14:21 00:54:06 1959333 318 270 0 48 +435.25 +/- 50.05
190140 ncm-et-3 2023-06-01 14:19 00:55:37 1962151 338 285 0 53 +428.08 +/- 47.51
190139 ncm-et-15 2023-06-01 14:18 00:56:37 1961537 338 290 0 48 +446.68 +/- 50.09
190138 ncm-et-9 2023-06-01 14:18 00:56:42 1949661 332 272 3 57 +391.81 +/- 45.88
190137 ncm-et-13 2023-06-01 14:14 01:00:36 1957418 359 304 1 54 +429.07 +/- 47.18
190130 ncm-et-4 2023-06-01 12:54 01:26:15 1953744 500 432 1 67 +452.04 +/- 42.22
190129 ncm-et-10 2023-06-01 12:54 01:26:17 1948006 500 422 0 78 +429.05 +/- 38.92
190128 ncm-et-3 2023-06-01 12:52 01:26:31 1940156 500 427 1 72 +438.95 +/- 40.66
190127 ncm-et-15 2023-06-01 12:52 01:25:37 1954805 500 428 3 69 +436.43 +/- 41.61
190126 ncm-et-9 2023-06-01 12:52 01:25:29 1952250 500 437 0 63 +468.95 +/- 43.54
190125 ncm-et-13 2023-06-01 12:48 01:25:59 1950019 500 425 3 72 +429.05 +/- 40.7
190118 ncm-et-10 2023-06-01 11:29 01:24:25 1960618 500 423 0 77 +431.48 +/- 39.18
190117 ncm-et-4 2023-06-01 11:28 01:25:39 1954865 500 440 1 59 +474.93 +/- 45.13
190116 ncm-et-15 2023-06-01 11:27 01:24:05 1960486 500 434 0 66 +460.32 +/- 42.48
190115 ncm-et-3 2023-06-01 11:26 01:25:34 1955600 500 422 0 78 +429.05 +/- 38.92
190114 ncm-et-9 2023-06-01 11:25 01:26:20 1956350 500 440 1 59 +474.93 +/- 45.13
190113 ncm-et-13 2023-06-01 11:21 01:26:14 1961694 500 427 0 73 +441.5 +/- 40.29
190106 ncm-et-10 2023-06-01 10:03 01:24:43 1946773 500 431 0 69 +452.04 +/- 41.5
190105 ncm-et-4 2023-06-01 10:02 01:25:42 1959870 500 441 0 59 +481.09 +/- 45.07
190104 ncm-et-15 2023-06-01 10:01 01:25:40 1917110 500 439 0 61 +474.93 +/- 44.28
190103 ncm-et-9 2023-06-01 10:01 01:23:32 1960169 500 425 2 73 +431.48 +/- 40.4
190102 ncm-et-3 2023-06-01 10:00 01:24:58 1945197 500 426 0 74 +438.95 +/- 40.01
190101 ncm-et-13 2023-06-01 09:57 01:22:51 1962948 500 419 2 79 +417.32 +/- 38.77
190094 ncm-et-10 2023-06-01 08:36 01:26:39 1950721 500 426 0 74 +438.95 +/- 40.01
190093 ncm-et-15 2023-06-01 08:35 01:25:13 1947969 500 426 1 73 +436.43 +/- 40.36
190092 ncm-et-3 2023-06-01 08:35 01:24:28 1961997 500 437 1 62 +466.04 +/- 43.97
190091 ncm-et-4 2023-06-01 08:35 01:25:56 1957726 500 437 0 63 +468.95 +/- 43.54
190090 ncm-et-9 2023-06-01 08:34 01:26:10 1939254 500 451 0 49 +515.19 +/- 49.71
190089 ncm-et-13 2023-06-01 08:33 01:23:20 1963876 500 427 3 70 +433.94 +/- 41.3
190082 ncm-et-3 2023-06-01 07:10 01:24:03 1960184 500 427 0 73 +441.5 +/- 40.29
190081 ncm-et-10 2023-06-01 07:10 01:25:35 1957271 500 430 1 69 +446.7 +/- 41.57
190080 ncm-et-15 2023-06-01 07:09 01:25:16 1959145 500 432 1 67 +452.04 +/- 42.22
190079 ncm-et-9 2023-06-01 07:09 01:24:39 1955873 500 433 0 67 +457.52 +/- 42.15
190078 ncm-et-13 2023-06-01 07:09 01:24:02 1959301 500 437 1 62 +466.04 +/- 43.97
190077 ncm-et-4 2023-06-01 07:08 01:25:55 1937101 500 426 0 74 +438.95 +/- 40.01
190070 ncm-et-15 2023-06-01 05:44 01:24:39 1956495 500 438 0 62 +471.92 +/- 43.9
190069 ncm-et-4 2023-06-01 05:44 01:23:48 1965226 500 436 1 63 +463.16 +/- 43.6
190068 ncm-et-10 2023-06-01 05:44 01:25:20 1950393 500 435 1 64 +460.32 +/- 43.24
190067 ncm-et-9 2023-06-01 05:44 01:24:27 1960487 500 436 2 62 +460.32 +/- 43.99
190066 ncm-et-3 2023-06-01 05:44 01:26:12 1954829 500 434 1 65 +457.52 +/- 42.89
190065 ncm-et-13 2023-06-01 05:43 01:24:28 1959405 500 432 0 68 +454.76 +/- 41.82

Commit

Commit ID c1fff71650e2f8bf5a2d63bdc043161cdfe8e460
Author Linmiao Xu
Date 2023-05-31 06:51:22 UTC
Update NNUE architecture to SFNNv6 with larger L1 size of 1536 Created by training a new net from scratch with L1 size increased from 1024 to 1536. Thanks to Vizvezdenec for the idea of exploring larger net sizes after recent training data improvements. A new net was first trained with lambda 1.0 and constant LR 8.75e-4. Then a strong net from a later epoch in the training run was chosen for retraining with start-lambda 1.0 and initial LR 4.375e-4 decaying with gamma 0.995. Retraining was performed a total of 3 times, for this 4-step process: 1. 400 epochs, lambda 1.0 on filtered T77+T79 v6 deduplicated data 2. 800 epochs, end-lambda 0.75 on T60T70wIsRightFarseerT60T74T75T76.binpack 3. 800 epochs, end-lambda 0.75 and early-fen-skipping 28 on the master dataset 4. 800 epochs, end-lambda 0.7 and early-fen-skipping 28 on the master dataset In the training sequence that reached the new nn-8d69132723e2.nnue net, the epochs used for the 3x retraining runs were: 1. epoch 379 trained on T77T79-filter-v6-dd.min.binpack 2. epoch 679 trained on T60T70wIsRightFarseerT60T74T75T76.binpack 3. epoch 799 trained on the master dataset For training from scratch: python3 easy_train.py \ --experiment-name new-L1-1536-T77T79-filter-v6dd \ --training-dataset /data/T77T79-filter-v6-dd.min.binpack \ --max_epoch 400 \ --lambda 1.0 \ --start-from-engine-test-net False \ --engine-test-branch linrock/Stockfish/L1-1536 \ --nnue-pytorch-branch linrock/Stockfish/misc-fixes-L1-1536 \ --tui False \ --gpus "0," \ --seed $RANDOM Retraining commands were similar to each other. For the 3rd retraining run: python3 easy_train.py \ --experiment-name L1-1536-T77T79-v6dd-Re1-LeelaFarseer-Re2-masterDataset-Re3-sameData \ --training-dataset /data/leela96-dfrc99-v2-T60novdecT80juntonovjanfebT79aprmayT78jantosepT77dec-v6dd.binpack \ --early-fen-skipping 28 \ --max_epoch 800 \ --start-lambda 1.0 \ --end-lambda 0.7 \ --lr 4.375e-4 \ --gamma 0.995 \ --start-from-engine-test-net False \ --start-from-model /data/L1-1536-T77T79-v6dd-Re1-LeelaFarseer-Re2-masterDataset-nn-epoch799.nnue \ --engine-test-branch linrock/Stockfish/L1-1536 \ --nnue-pytorch-branch linrock/nnue-pytorch/misc-fixes-L1-1536 \ --tui False \ --gpus "0," \ --seed $RANDOM The T77+T79 data used is a subset of the master dataset available at: https://robotmoon.com/nnue-training-data/ T60T70wIsRightFarseerT60T74T75T76.binpack is available at: https://drive.google.com/drive/folders/1S9-ZiQa_3ApmjBtl2e8SyHxj4zG4V8gG Local elo at 25k nodes per move vs. nn-e1fb1ade4432.nnue (L1 size 1024): nn-epoch759.nnue : 26.9 +/- 1.6 Failed STC https://tests.stockfishchess.org/tests/view/64742485d29264e4cfa75f97 LLR: -2.94 (-2.94,2.94) <0.00,2.00> Total: 13728 W: 3588 L: 3829 D: 6311 Ptnml(0-2): 71, 1661, 3610, 1482, 40 Failing LTC https://tests.stockfishchess.org/tests/view/64752d7c4a36543c4c9f3618 LLR: -1.91 (-2.94,2.94) <0.50,2.50> Total: 35424 W: 9522 L: 9603 D: 16299 Ptnml(0-2): 24, 3579, 10585, 3502, 22 Passed VLTC 180+1.8 https://tests.stockfishchess.org/tests/view/64752df04a36543c4c9f3638 LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 47616 W: 13174 L: 12863 D: 21579 Ptnml(0-2): 13, 4261, 14952, 4566, 16 Passed VLTC SMP 60+0.6 th 8 https://tests.stockfishchess.org/tests/view/647446ced29264e4cfa761e5 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 19942 W: 5694 L: 5451 D: 8797 Ptnml(0-2): 6, 1504, 6707, 1749, 5 closes https://github.com/official-stockfish/Stockfish/pull/4593 bench 2222567
Copyright 2011–2024 Next Chess Move LLC