Buckets:
| 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.