patrickfrank1's picture
download
raw
19.3 kB
device=cuda params=3.19M bf16=True compile=False
frequency-aware loss weighting enabled (streaming counts)
train games=651466 val games=34396
step=50 loss=3.7738 lr=3.00e-05 tokens/s=495212
step=100 loss=3.0315 lr=6.00e-05 tokens/s=967852
step=150 loss=2.7800 lr=9.00e-05 tokens/s=1011521
step=200 loss=2.6918 lr=1.20e-04 tokens/s=1000609
step=250 loss=2.6494 lr=1.50e-04 tokens/s=1002540
step=300 loss=2.6227 lr=1.80e-04 tokens/s=1008820
step=350 loss=2.6122 lr=2.10e-04 tokens/s=1012771
step=400 loss=2.5872 lr=2.40e-04 tokens/s=990883
step=450 loss=2.5810 lr=2.70e-04 tokens/s=1004819
step=500 loss=2.5549 lr=3.00e-04 tokens/s=998729
step=550 loss=2.5568 lr=3.00e-04 tokens/s=975301
step=600 loss=2.5448 lr=3.00e-04 tokens/s=991336
step=650 loss=2.5342 lr=3.00e-04 tokens/s=996955
step=700 loss=2.5313 lr=3.00e-04 tokens/s=987162
step=750 loss=2.5266 lr=3.00e-04 tokens/s=999528
step=800 loss=2.5211 lr=2.99e-04 tokens/s=432634
step=850 loss=2.5109 lr=2.99e-04 tokens/s=1017613
step=900 loss=2.5160 lr=2.99e-04 tokens/s=1012591
step=950 loss=2.5169 lr=2.99e-04 tokens/s=1004643
step=1000 loss=2.5158 lr=2.99e-04 tokens/s=206749
step=1000 val_loss=2.4778 val_acc=0.2615
step=1000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=1.000 material=+0 8/1p2kpn1/2p5/2P2pPp/1PB2P1P/4K3/8/8 b - - 0 1
step=1000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 1r1r1nk1/pBq1bpp1/Pp1ppn1p/2p5/QP1P4/2P1PN1P/5PPB/3R1RK1 b - - 0 1
step=1000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.989 material=-1 8/5p2/6k1/2PPK3/6pp/4n1P1/2r2P1P/R7 w - - 0 1
step=1000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=1000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=1000 board_turn_acc=0.1375 board_castle_acc=0.2625 board_piece_acc=0.3223 board_prior_baseline=0.5947 random_piece_acc=0.2832 span_piece_acc=0.3405 cos_to_move_child=0.9998 cos_to_corrupted=0.9932 test_r2_material=0.6274 pc1_piece_corr=-0.9716
saved best checkpoint to data/models/tiny_d256/best.pt (val_loss=2.4778)
uploaded best_step1000 to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=1050 loss=2.5072 lr=2.98e-04 tokens/s=28498
step=1100 loss=2.5100 lr=2.98e-04 tokens/s=125619
step=1150 loss=2.4999 lr=2.98e-04 tokens/s=988933
step=1200 loss=2.4996 lr=2.97e-04 tokens/s=136059
step=1250 loss=2.4967 lr=2.97e-04 tokens/s=1007394
step=1300 loss=2.4911 lr=2.96e-04 tokens/s=535554
step=1350 loss=2.4988 lr=2.96e-04 tokens/s=983893
step=1400 loss=2.4729 lr=2.95e-04 tokens/s=979879
step=1450 loss=2.4343 lr=2.95e-04 tokens/s=989483
step=1500 loss=2.4335 lr=2.94e-04 tokens/s=988546
step=1550 loss=2.4346 lr=2.94e-04 tokens/s=983482
step=1600 loss=2.4321 lr=2.93e-04 tokens/s=1000578
step=1650 loss=2.4191 lr=2.93e-04 tokens/s=998891
step=1700 loss=2.4183 lr=2.92e-04 tokens/s=436744
step=1750 loss=2.4337 lr=2.91e-04 tokens/s=982724
step=1800 loss=2.4169 lr=2.91e-04 tokens/s=986026
step=1850 loss=2.4181 lr=2.90e-04 tokens/s=36369
step=1900 loss=2.4147 lr=2.89e-04 tokens/s=995757
step=1950 loss=2.4188 lr=2.88e-04 tokens/s=985765
step=2000 loss=2.4246 lr=2.88e-04 tokens/s=977410
step=2000 val_loss=2.5575 val_acc=0.1326
step=2000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.999 material=+0 8/1p2kpn1/2p5/2P2pPp/1PB2P1P/4K3/8/8 b - - 0 1
step=2000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 1r1r1nk1/pBq1bpp1/Pp1ppn1p/2p5/QP1P4/2P1PN1P/5PPB/3R1RK1 b - - 0 1
step=2000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.998 material=+2 8/3B4/2Pk4/p5p1/P3KnP1/5P2/8/8 b - - 0 1
step=2000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=2000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=2000 board_turn_acc=0.8125 board_castle_acc=0.3875 board_piece_acc=0.1201 board_prior_baseline=0.5187 random_piece_acc=0.1260 span_piece_acc=0.1240 cos_to_move_child=0.9998 cos_to_corrupted=0.9990 test_r2_material=0.4245 pc1_piece_corr=-0.9896
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=2050 loss=2.4240 lr=2.87e-04 tokens/s=43775
step=2100 loss=2.4047 lr=2.86e-04 tokens/s=1006892
step=2150 loss=2.4062 lr=2.85e-04 tokens/s=734837
step=2200 loss=2.3998 lr=2.84e-04 tokens/s=988067
step=2250 loss=2.3634 lr=2.83e-04 tokens/s=999860
step=2300 loss=2.3734 lr=2.82e-04 tokens/s=996727
step=2350 loss=2.3675 lr=2.81e-04 tokens/s=1014068
step=2400 loss=2.3525 lr=2.80e-04 tokens/s=1015084
step=2450 loss=2.3446 lr=2.79e-04 tokens/s=999870
step=2500 loss=2.3449 lr=2.78e-04 tokens/s=994650
step=2550 loss=2.3300 lr=2.77e-04 tokens/s=1008447
step=2600 loss=2.3326 lr=2.76e-04 tokens/s=1007118
step=2650 loss=2.3283 lr=2.75e-04 tokens/s=997459
step=2700 loss=2.3267 lr=2.74e-04 tokens/s=512340
step=2750 loss=2.3231 lr=2.73e-04 tokens/s=1011292
step=2800 loss=2.3289 lr=2.71e-04 tokens/s=1017686
step=2850 loss=2.3212 lr=2.70e-04 tokens/s=1012214
step=2900 loss=2.3249 lr=2.69e-04 tokens/s=1004964
step=2950 loss=2.3233 lr=2.68e-04 tokens/s=1005177
step=3000 loss=2.3210 lr=2.66e-04 tokens/s=997786
step=3000 val_loss=2.5151 val_acc=0.1579
step=3000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.999 material=+0 8/1p2kpn1/2p5/2P2pPp/1PB2P1P/4K3/8/8 b - - 0 1
step=3000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 rn3rk1/ppq1ppbp/2pp1np1/8/2PP4/1QN1PB2/PP1B1PPP/R4RK1 b - - 0 1
step=3000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.998 material=+2 8/3B4/2Pk4/p5p1/P3KnP1/5P2/8/8 b - - 0 1
step=3000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=3000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=3000 board_turn_acc=0.8375 board_castle_acc=0.4750 board_piece_acc=0.1252 board_prior_baseline=0.5194 random_piece_acc=0.1972 span_piece_acc=0.1208 cos_to_move_child=0.9991 cos_to_corrupted=0.9980 test_r2_material=0.4733 pc1_piece_corr=-0.9740
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=3050 loss=2.3153 lr=2.65e-04 tokens/s=24144
step=3100 loss=2.3291 lr=2.64e-04 tokens/s=1008295
step=3150 loss=2.3148 lr=2.62e-04 tokens/s=1009573
step=3200 loss=2.3138 lr=2.61e-04 tokens/s=1002112
step=3250 loss=2.3084 lr=2.60e-04 tokens/s=999564
step=3300 loss=2.3104 lr=2.58e-04 tokens/s=997285
step=3350 loss=2.2979 lr=2.57e-04 tokens/s=1003716
step=3400 loss=2.3165 lr=2.55e-04 tokens/s=995078
step=3450 loss=2.3074 lr=2.54e-04 tokens/s=995896
step=3500 loss=2.3132 lr=2.52e-04 tokens/s=981282
step=3550 loss=2.3110 lr=2.51e-04 tokens/s=999250
step=3600 loss=2.3099 lr=2.49e-04 tokens/s=996484
step=3650 loss=2.3157 lr=2.48e-04 tokens/s=1011407
step=3700 loss=2.3065 lr=2.46e-04 tokens/s=1014087
step=3750 loss=2.3111 lr=2.45e-04 tokens/s=1001047
step=3800 loss=2.3138 lr=2.43e-04 tokens/s=1011417
step=3850 loss=2.3008 lr=2.41e-04 tokens/s=1016437
step=3900 loss=2.3041 lr=2.40e-04 tokens/s=1008113
step=3950 loss=2.3012 lr=2.38e-04 tokens/s=1001178
step=4000 loss=2.2999 lr=2.37e-04 tokens/s=980634
step=4000 val_loss=2.5228 val_acc=0.1566
step=4000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.999 material=+0 8/1p2kpn1/2p5/2P2pPp/1PB2P1P/4K3/8/8 b - - 0 1
step=4000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 1r1r1nk1/pBq1bpp1/Pp1ppn1p/2p5/QP1P4/2P1PN1P/5PPB/3R1RK1 b - - 0 1
step=4000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=1.000 material=-1 8/6R1/8/4p1pp/4P2k/5P2/5K2/3r4 b - - 0 1
step=4000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=4000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=4000 board_turn_acc=0.8625 board_castle_acc=0.4875 board_piece_acc=0.1238 board_prior_baseline=0.5187 random_piece_acc=0.2072 span_piece_acc=0.1195 cos_to_move_child=0.9995 cos_to_corrupted=0.9953 test_r2_material=0.4537 pc1_piece_corr=0.9778
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=4050 loss=2.2825 lr=2.35e-04 tokens/s=20290
step=4100 loss=2.2983 lr=2.33e-04 tokens/s=1019193
step=4150 loss=2.2860 lr=2.31e-04 tokens/s=1002750
step=4200 loss=2.2933 lr=2.30e-04 tokens/s=985702
step=4250 loss=2.2908 lr=2.28e-04 tokens/s=1001372
step=4300 loss=2.2855 lr=2.26e-04 tokens/s=1015463
step=4350 loss=2.2825 lr=2.24e-04 tokens/s=1001091
step=4400 loss=2.2962 lr=2.23e-04 tokens/s=1007394
step=4450 loss=2.2761 lr=2.21e-04 tokens/s=1007245
step=4500 loss=2.2822 lr=2.19e-04 tokens/s=1019325
step=4550 loss=2.2557 lr=2.17e-04 tokens/s=1019014
step=4600 loss=2.2789 lr=2.15e-04 tokens/s=1000779
step=4650 loss=2.2529 lr=2.13e-04 tokens/s=1015899
step=4700 loss=2.2642 lr=2.12e-04 tokens/s=1001591
step=4750 loss=2.2567 lr=2.10e-04 tokens/s=983403
step=4800 loss=2.2760 lr=2.08e-04 tokens/s=1010639
step=4850 loss=2.2438 lr=2.06e-04 tokens/s=991639
step=4900 loss=2.2712 lr=2.04e-04 tokens/s=1001907
step=4950 loss=2.2517 lr=2.02e-04 tokens/s=1008506
step=5000 loss=2.2564 lr=2.00e-04 tokens/s=1009893
step=5000 val_loss=2.5130 val_acc=0.1613
step=5000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.999 material=-7 5r2/7p/1b1k2pP/2pP2PB/2Pp1p2/p4K2/8/8 w - - 0 1
step=5000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 rn3rk1/ppq1ppbp/2pp1np1/8/2PP4/1QN1PB2/PP1B1PPP/R4RK1 b - - 0 1
step=5000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.999 material=-1 8/6R1/8/4p1pp/4P2k/5P2/5K2/3r4 b - - 0 1
step=5000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=5000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=5000 board_turn_acc=0.8500 board_castle_acc=0.4875 board_piece_acc=0.1154 board_prior_baseline=0.5161 random_piece_acc=0.2112 span_piece_acc=0.1223 cos_to_move_child=0.9994 cos_to_corrupted=0.9744 test_r2_material=0.4967 pc1_piece_corr=0.9521
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=5050 loss=2.2596 lr=1.98e-04 tokens/s=34746
step=5100 loss=2.2665 lr=1.96e-04 tokens/s=993378
step=5150 loss=2.2525 lr=1.94e-04 tokens/s=1002906
step=5200 loss=2.2530 lr=1.92e-04 tokens/s=1006679
step=5250 loss=2.2433 lr=1.90e-04 tokens/s=1005912
step=5300 loss=2.2567 lr=1.88e-04 tokens/s=1022962
step=5350 loss=2.2417 lr=1.87e-04 tokens/s=1021528
step=5400 loss=2.2578 lr=1.85e-04 tokens/s=1017130
step=5450 loss=2.2456 lr=1.83e-04 tokens/s=1008746
step=5500 loss=2.2515 lr=1.81e-04 tokens/s=1003531
step=5550 loss=2.2369 lr=1.79e-04 tokens/s=987692
step=5600 loss=2.2506 lr=1.76e-04 tokens/s=988715
step=5650 loss=2.2440 lr=1.74e-04 tokens/s=1009504
step=5700 loss=2.2574 lr=1.72e-04 tokens/s=1012478
step=5750 loss=2.2371 lr=1.70e-04 tokens/s=1003165
step=5800 loss=2.2526 lr=1.68e-04 tokens/s=991608
step=5850 loss=2.2389 lr=1.66e-04 tokens/s=992038
step=5900 loss=2.2524 lr=1.64e-04 tokens/s=985384
step=5950 loss=2.2399 lr=1.62e-04 tokens/s=991403
step=6000 loss=2.2528 lr=1.60e-04 tokens/s=1001048
step=6000 val_loss=2.4601 val_acc=0.1636
step=6000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.999 material=+0 8/1p2kpn1/2p5/2P2pPp/1PB2P1P/4K3/8/8 b - - 0 1
step=6000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 rn3rk1/ppq1ppbp/2pp1np1/8/2PP4/1QN1PB2/PP1B1PPP/R4RK1 b - - 0 1
step=6000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.999 material=-1 8/6R1/8/4p1pp/4P2k/5P2/5K2/3r4 b - - 0 1
step=6000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=6000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=6000 board_turn_acc=0.8375 board_castle_acc=0.4625 board_piece_acc=0.1262 board_prior_baseline=0.5199 random_piece_acc=0.2244 span_piece_acc=0.1215 cos_to_move_child=0.9992 cos_to_corrupted=0.9785 test_r2_material=0.5889 pc1_piece_corr=-0.9258
saved best checkpoint to data/models/tiny_d256/best.pt (val_loss=2.4601)
uploaded best_step6000 to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=6050 loss=2.2515 lr=1.58e-04 tokens/s=33232
step=6100 loss=2.2539 lr=1.56e-04 tokens/s=1001306
step=6150 loss=2.2467 lr=1.54e-04 tokens/s=1005842
step=6200 loss=2.2489 lr=1.52e-04 tokens/s=952588
step=6250 loss=2.2303 lr=1.50e-04 tokens/s=970979
step=6300 loss=2.2211 lr=1.48e-04 tokens/s=994117
step=6350 loss=2.2371 lr=1.46e-04 tokens/s=996128
step=6400 loss=2.2402 lr=1.44e-04 tokens/s=1000475
step=6450 loss=2.2278 lr=1.42e-04 tokens/s=988576
step=6500 loss=2.2435 lr=1.40e-04 tokens/s=993875
step=6550 loss=2.2372 lr=1.38e-04 tokens/s=992627
step=6600 loss=2.2390 lr=1.36e-04 tokens/s=997195
step=6650 loss=2.2328 lr=1.34e-04 tokens/s=987820
step=6700 loss=2.2384 lr=1.32e-04 tokens/s=998116
step=6750 loss=2.2124 lr=1.30e-04 tokens/s=993668
step=6800 loss=2.2302 lr=1.28e-04 tokens/s=977924
step=6850 loss=2.2276 lr=1.26e-04 tokens/s=977869
step=6900 loss=2.2277 lr=1.24e-04 tokens/s=997859
step=6950 loss=2.2459 lr=1.21e-04 tokens/s=1015831
step=7000 loss=2.2311 lr=1.19e-04 tokens/s=1004475
step=7000 val_loss=2.4218 val_acc=0.1725
step=7000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.999 material=+0 8/1p2kpn1/2p5/2P2pPp/1PB2P1P/4K3/8/8 b - - 0 1
step=7000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 rn3rk1/ppq1ppbp/2pp1np1/8/2PP4/1QN1PB2/PP1B1PPP/R4RK1 b - - 0 1
step=7000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.996 material=+3 8/8/3N1kp1/3n3p/8/6KP/5NP1/8 b - - 0 1
step=7000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=7000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=7000 board_turn_acc=0.8500 board_castle_acc=0.5000 board_piece_acc=0.1275 board_prior_baseline=0.5200 random_piece_acc=0.2524 span_piece_acc=0.1240 cos_to_move_child=0.9994 cos_to_corrupted=0.9760 test_r2_material=0.6864 pc1_piece_corr=-0.9089
saved best checkpoint to data/models/tiny_d256/best.pt (val_loss=2.4218)
uploaded best_step7000 to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=7050 loss=2.2167 lr=1.17e-04 tokens/s=24433
step=7100 loss=2.2411 lr=1.15e-04 tokens/s=1006032
step=7150 loss=2.2224 lr=1.13e-04 tokens/s=507576
step=7200 loss=2.2264 lr=1.12e-04 tokens/s=1001325
step=7250 loss=2.2259 lr=1.10e-04 tokens/s=995404
step=7300 loss=2.2339 lr=1.08e-04 tokens/s=622096
step=7350 loss=2.2315 lr=1.06e-04 tokens/s=994206
step=7400 loss=2.2364 lr=1.04e-04 tokens/s=995250
step=7450 loss=2.2182 lr=1.02e-04 tokens/s=1016207
step=7500 loss=2.2252 lr=9.98e-05 tokens/s=426211
step=7550 loss=2.2283 lr=9.78e-05 tokens/s=1011740
step=7600 loss=2.2371 lr=9.59e-05 tokens/s=1010255
step=7650 loss=2.2173 lr=9.40e-05 tokens/s=1019879
step=7700 loss=2.2243 lr=9.21e-05 tokens/s=1004582
step=7750 loss=2.2324 lr=9.02e-05 tokens/s=983561
step=7800 loss=2.2156 lr=8.84e-05 tokens/s=1005255
step=7850 loss=2.2162 lr=8.65e-05 tokens/s=988705
step=7900 loss=2.2274 lr=8.47e-05 tokens/s=993941
step=7950 loss=2.2169 lr=8.28e-05 tokens/s=1001902
step=8000 loss=2.2343 lr=8.10e-05 tokens/s=1004520
step=8000 val_loss=2.4419 val_acc=0.1678
step=8000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.999 material=+0 8/1p2kpn1/2p5/2P2pPp/1PB2P1P/4K3/8/8 b - - 0 1
step=8000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 1r1r1nk1/pBq1bpp1/Pp1ppn1p/2p5/QP1P4/2P1PN1P/5PPB/3R1RK1 b - - 0 1
step=8000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.995 material=+0 4R3/8/2kNp1p1/2P3P1/3n1PK1/1r6/4p3/8 b - - 0 1
step=8000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=8000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=8000 board_turn_acc=0.8375 board_castle_acc=0.5125 board_piece_acc=0.1283 board_prior_baseline=0.5205 random_piece_acc=0.2360 span_piece_acc=0.1245 cos_to_move_child=0.9986 cos_to_corrupted=0.9725 test_r2_material=0.7024 pc1_piece_corr=-0.8796
saved full checkpoint to data/models/tiny_d256/last.pt
uploaded last to patrickfrank1/chesspos-checkpoints/run_tiny_d256_streaming
step=8050 loss=2.2147 lr=7.92e-05 tokens/s=45090
step=8100 loss=2.2123 lr=7.74e-05 tokens/s=1013892
step=8150 loss=2.2182 lr=7.56e-05 tokens/s=906652
step=8200 loss=2.2301 lr=7.38e-05 tokens/s=1012337
step=8250 loss=2.2145 lr=7.21e-05 tokens/s=992175
step=8300 loss=2.2230 lr=7.03e-05 tokens/s=1005069
step=8350 loss=2.2183 lr=6.86e-05 tokens/s=423905
step=8400 loss=2.2188 lr=6.69e-05 tokens/s=997604
step=8450 loss=2.2046 lr=6.52e-05 tokens/s=998316
step=8500 loss=2.2168 lr=6.35e-05 tokens/s=999787
step=8550 loss=2.2253 lr=6.18e-05 tokens/s=1003056
step=8600 loss=2.2162 lr=6.02e-05 tokens/s=995978
step=8650 loss=2.2110 lr=5.86e-05 tokens/s=1000001
step=8700 loss=2.2204 lr=5.69e-05 tokens/s=996524
step=8750 loss=2.2079 lr=5.53e-05 tokens/s=993504
step=8800 loss=2.2324 lr=5.38e-05 tokens/s=1007902
step=8850 loss=2.1961 lr=5.22e-05 tokens/s=130176
step=8900 loss=2.2111 lr=5.06e-05 tokens/s=1017266
step=8950 loss=2.2167 lr=4.91e-05 tokens/s=1018596
step=9000 loss=2.2193 lr=4.76e-05 tokens/s=1019700
step=9000 val_loss=2.4246 val_acc=0.1684
step=9000 neighbor query material=-1 8/p4p2/4k2p/1pp1P1p1/3n1PP1/3BK2P/P7/8 w - - 0 1
top sim=0.998 material=+1 8/8/p1k1nppp/1p6/2p1PPPP/2P1K3/PPB5/8 w - - 0 1
step=9000 neighbor query material=+0 rn1q1rk1/pbp2pp1/1p1ppn1p/8/2PP3B/2P1PN2/P3BPPP/R2Q1RK1 b - - 0 1
top sim=1.000 material=+0 rn3rk1/ppq1ppbp/2pp1np1/8/2PP4/1QN1PB2/PP1B1PPP/R4RK1 b - - 0 1
step=9000 neighbor query material=-7 8/6pk/7p/8/n3P3/4K3/1r3P1P/8 b - - 0 1
top sim=0.991 material=+0 4R3/8/2kNp1p1/2P3P1/3n1PK1/1r6/4p3/8 b - - 0 1
step=9000 neighbor board report saved to data/models/tiny_d256/neighbor_examples.html
step=9000 eval plot saved to data/models/tiny_d256/pca_phase.png
step=9000 board_turn_acc=0.8375 board_castle_acc=0.5000 board_piece_acc=0.1287 board_prior_baseline=0.5209 random_piece_acc=0.2368 span_piece_acc=0.1245 cos_to_move_child=0.9979 cos_to_corrupted=0.9701 test_r2_material=0.7437 pc1_piece_corr=-0.8885
saved full checkpoint to data/models/tiny_d256/last.pt

Xet Storage Details

Size:
19.3 kB
·
Xet hash:
ad0ef13eee3bb3d85a4e2b1e3115bdc6389c5bac23fd74dbfca1224886ce8ff8

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.