Minesweeper RL Agent
Just for fun, on some compute:
(kits) z5362216@k108:~/minesweeper $ nvidia-smi
Tue Jan 13 21:33:28 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.95.05 Driver Version: 580.95.05 CUDA Version: 13.0 |
+-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 Tesla V100-SXM2-32GB On | 00000000:AF:00.0 Off | 0 |
| N/A 44C P0 42W / 300W | 0MiB / 32768MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
Outputs
Beginner
(kits) z5362216@k105:~/minesweeper $ python train_probability.py --games 200000 --epochs 50 --difficulty all --out models
Device: cuda
============================================================
easy (9x9, 10 mines)
============================================================
Phase 1: Collecting training data from 200000 games...
Collected 1000 games, 1271 guess states, win rate: 85.2%
Collected 2000 games, 2511 guess states, win rate: 87.3%
Collected 3000 games, 3781 guess states, win rate: 87.5%
Collected 4000 games, 5031 guess states, win rate: 87.2%
Collected 5000 games, 6293 guess states, win rate: 87.4%
Collected 6000 games, 7540 guess states, win rate: 87.4%
Collected 7000 games, 8778 guess states, win rate: 87.3%
Collected 8000 games, 10018 guess states, win rate: 87.6%
Collected 9000 games, 11240 guess states, win rate: 87.7%
Collected 10000 games, 12458 guess states, win rate: 87.8%
Collected 11000 games, 13721 guess states, win rate: 87.7%
Collected 12000 games, 14982 guess states, win rate: 87.7%
Collected 13000 games, 16231 guess states, win rate: 87.8%
Collected 14000 games, 17428 guess states, win rate: 87.9%
Collected 15000 games, 18707 guess states, win rate: 87.9%
Collected 16000 games, 19980 guess states, win rate: 87.9%
Collected 17000 games, 21253 guess states, win rate: 87.8%
Collected 18000 games, 22544 guess states, win rate: 87.8%
Collected 19000 games, 23787 guess states, win rate: 87.8%
Collected 20000 games, 25018 guess states, win rate: 87.8%
Collected 21000 games, 26240 guess states, win rate: 87.9%
Collected 22000 games, 27475 guess states, win rate: 88.0%
Collected 23000 games, 28756 guess states, win rate: 87.9%
Collected 24000 games, 30056 guess states, win rate: 87.8%
Collected 25000 games, 31300 guess states, win rate: 87.8%
Collected 26000 games, 32574 guess states, win rate: 87.7%
Collected 27000 games, 33821 guess states, win rate: 87.7%
Collected 28000 games, 35035 guess states, win rate: 87.8%
Collected 29000 games, 36252 guess states, win rate: 87.8%
Collected 30000 games, 37475 guess states, win rate: 87.8%
Collected 31000 games, 38715 guess states, win rate: 87.8%
Collected 32000 games, 39959 guess states, win rate: 87.9%
Collected 33000 games, 41230 guess states, win rate: 87.9%
Collected 34000 games, 42478 guess states, win rate: 87.9%
Collected 35000 games, 43720 guess states, win rate: 87.9%
Collected 36000 games, 44967 guess states, win rate: 87.9%
Collected 37000 games, 46182 guess states, win rate: 87.9%
Collected 38000 games, 47398 guess states, win rate: 87.9%
Collected 39000 games, 48609 guess states, win rate: 88.0%
Collected 40000 games, 49850 guess states, win rate: 88.0%
Collected 41000 games, 51111 guess states, win rate: 88.0%
Collected 42000 games, 52342 guess states, win rate: 88.0%
Collected 43000 games, 53577 guess states, win rate: 88.0%
Collected 44000 games, 54794 guess states, win rate: 88.0%
Collected 45000 games, 56069 guess states, win rate: 88.0%
Collected 46000 games, 57299 guess states, win rate: 88.0%
Collected 47000 games, 58580 guess states, win rate: 88.0%
Collected 48000 games, 59807 guess states, win rate: 88.0%
Collected 49000 games, 61054 guess states, win rate: 88.0%
Collected 50000 games, 62278 guess states, win rate: 88.0%
Collected 51000 games, 63510 guess states, win rate: 88.0%
Collected 52000 games, 64720 guess states, win rate: 88.1%
Collected 53000 games, 65980 guess states, win rate: 88.1%
Collected 54000 games, 67191 guess states, win rate: 88.1%
Collected 55000 games, 68442 guess states, win rate: 88.1%
Collected 56000 games, 69706 guess states, win rate: 88.1%
Collected 57000 games, 70941 guess states, win rate: 88.1%
Collected 58000 games, 72194 guess states, win rate: 88.1%
Collected 59000 games, 73433 guess states, win rate: 88.1%
Collected 60000 games, 74644 guess states, win rate: 88.1%
Collected 61000 games, 75892 guess states, win rate: 88.1%
Collected 62000 games, 77140 guess states, win rate: 88.1%
Collected 63000 games, 78401 guess states, win rate: 88.1%
Collected 64000 games, 79642 guess states, win rate: 88.1%
Collected 65000 games, 80871 guess states, win rate: 88.1%
Collected 66000 games, 82113 guess states, win rate: 88.1%
Collected 67000 games, 83396 guess states, win rate: 88.1%
Collected 68000 games, 84645 guess states, win rate: 88.1%
Collected 69000 games, 85868 guess states, win rate: 88.1%
Collected 70000 games, 87081 guess states, win rate: 88.1%
Collected 71000 games, 88319 guess states, win rate: 88.1%
Collected 72000 games, 89565 guess states, win rate: 88.1%
Collected 73000 games, 90836 guess states, win rate: 88.1%
Collected 74000 games, 92050 guess states, win rate: 88.1%
Collected 75000 games, 93288 guess states, win rate: 88.1%
Collected 76000 games, 94512 guess states, win rate: 88.2%
Collected 77000 games, 95741 guess states, win rate: 88.2%
Collected 78000 games, 96983 guess states, win rate: 88.2%
Collected 79000 games, 98199 guess states, win rate: 88.2%
Collected 80000 games, 99442 guess states, win rate: 88.2%
Collected 81000 games, 100715 guess states, win rate: 88.2%
Collected 82000 games, 101942 guess states, win rate: 88.2%
Collected 83000 games, 103182 guess states, win rate: 88.2%
Collected 84000 games, 104416 guess states, win rate: 88.2%
Collected 85000 games, 105643 guess states, win rate: 88.2%
Collected 86000 games, 106866 guess states, win rate: 88.2%
Collected 87000 games, 108129 guess states, win rate: 88.2%
Collected 88000 games, 109340 guess states, win rate: 88.2%
Collected 89000 games, 110602 guess states, win rate: 88.2%
Collected 90000 games, 111853 guess states, win rate: 88.2%
Collected 91000 games, 113077 guess states, win rate: 88.1%
Collected 92000 games, 114305 guess states, win rate: 88.1%
Collected 93000 games, 115537 guess states, win rate: 88.1%
Collected 94000 games, 116809 guess states, win rate: 88.1%
Collected 95000 games, 118051 guess states, win rate: 88.1%
Collected 96000 games, 119286 guess states, win rate: 88.1%
Collected 97000 games, 120549 guess states, win rate: 88.1%
Collected 98000 games, 121832 guess states, win rate: 88.1%
Collected 99000 games, 123061 guess states, win rate: 88.1%
Collected 100000 games, 124285 guess states, win rate: 88.1%
Collected 101000 games, 125517 guess states, win rate: 88.1%
Collected 102000 games, 126739 guess states, win rate: 88.1%
Collected 103000 games, 127991 guess states, win rate: 88.1%
Collected 104000 games, 129214 guess states, win rate: 88.2%
Collected 105000 games, 130464 guess states, win rate: 88.2%
Collected 106000 games, 131706 guess states, win rate: 88.2%
Collected 107000 games, 132945 guess states, win rate: 88.2%
Collected 108000 games, 134200 guess states, win rate: 88.2%
Collected 109000 games, 135434 guess states, win rate: 88.1%
Collected 110000 games, 136671 guess states, win rate: 88.2%
Collected 111000 games, 137939 guess states, win rate: 88.1%
Collected 112000 games, 139148 guess states, win rate: 88.1%
Collected 113000 games, 140399 guess states, win rate: 88.1%
Collected 114000 games, 141650 guess states, win rate: 88.1%
Collected 115000 games, 142893 guess states, win rate: 88.1%
Collected 116000 games, 144116 guess states, win rate: 88.1%
Collected 117000 games, 145343 guess states, win rate: 88.1%
Collected 118000 games, 146564 guess states, win rate: 88.1%
Collected 119000 games, 147827 guess states, win rate: 88.1%
Collected 120000 games, 149077 guess states, win rate: 88.2%
Collected 121000 games, 150287 guess states, win rate: 88.2%
Collected 122000 games, 151538 guess states, win rate: 88.2%
Collected 123000 games, 152769 guess states, win rate: 88.2%
Collected 124000 games, 153978 guess states, win rate: 88.2%
Collected 125000 games, 155220 guess states, win rate: 88.2%
Collected 126000 games, 156488 guess states, win rate: 88.2%
Collected 127000 games, 157724 guess states, win rate: 88.2%
Collected 128000 games, 158976 guess states, win rate: 88.2%
Collected 129000 games, 160228 guess states, win rate: 88.2%
Collected 130000 games, 161418 guess states, win rate: 88.2%
Collected 131000 games, 162636 guess states, win rate: 88.2%
Collected 132000 games, 163857 guess states, win rate: 88.2%
Collected 133000 games, 165098 guess states, win rate: 88.2%
Collected 134000 games, 166342 guess states, win rate: 88.2%
Collected 135000 games, 167597 guess states, win rate: 88.2%
Collected 136000 games, 168844 guess states, win rate: 88.2%
Collected 137000 games, 170081 guess states, win rate: 88.2%
Collected 138000 games, 171361 guess states, win rate: 88.2%
Collected 139000 games, 172619 guess states, win rate: 88.2%
Collected 140000 games, 173838 guess states, win rate: 88.2%
Collected 141000 games, 175084 guess states, win rate: 88.2%
Collected 142000 games, 176315 guess states, win rate: 88.2%
Collected 143000 games, 177589 guess states, win rate: 88.2%
Collected 144000 games, 178865 guess states, win rate: 88.2%
Collected 145000 games, 180095 guess states, win rate: 88.2%
Collected 146000 games, 181329 guess states, win rate: 88.2%
Collected 147000 games, 182571 guess states, win rate: 88.2%
Collected 148000 games, 183795 guess states, win rate: 88.2%
Collected 149000 games, 185032 guess states, win rate: 88.2%
Collected 150000 games, 186263 guess states, win rate: 88.2%
Collected 151000 games, 187523 guess states, win rate: 88.2%
Collected 152000 games, 188753 guess states, win rate: 88.2%
Collected 153000 games, 189992 guess states, win rate: 88.2%
Collected 154000 games, 191210 guess states, win rate: 88.2%
Collected 155000 games, 192456 guess states, win rate: 88.2%
Collected 156000 games, 193709 guess states, win rate: 88.2%
Collected 157000 games, 194939 guess states, win rate: 88.2%
Collected 158000 games, 196163 guess states, win rate: 88.2%
Collected 159000 games, 197436 guess states, win rate: 88.2%
Collected 160000 games, 198656 guess states, win rate: 88.2%
Collected 161000 games, 199903 guess states, win rate: 88.2%
Collected 162000 games, 201103 guess states, win rate: 88.2%
Collected 163000 games, 202334 guess states, win rate: 88.2%
Collected 164000 games, 203571 guess states, win rate: 88.2%
Collected 165000 games, 204831 guess states, win rate: 88.2%
Collected 166000 games, 206066 guess states, win rate: 88.2%
Collected 167000 games, 207305 guess states, win rate: 88.2%
Collected 168000 games, 208528 guess states, win rate: 88.2%
Collected 169000 games, 209811 guess states, win rate: 88.2%
Collected 170000 games, 211030 guess states, win rate: 88.2%
Collected 171000 games, 212288 guess states, win rate: 88.2%
Collected 172000 games, 213502 guess states, win rate: 88.2%
Collected 173000 games, 214730 guess states, win rate: 88.2%
Collected 174000 games, 215967 guess states, win rate: 88.2%
Collected 175000 games, 217199 guess states, win rate: 88.2%
Collected 176000 games, 218456 guess states, win rate: 88.2%
Collected 177000 games, 219682 guess states, win rate: 88.2%
Collected 178000 games, 220934 guess states, win rate: 88.2%
Collected 179000 games, 222186 guess states, win rate: 88.2%
Collected 180000 games, 223438 guess states, win rate: 88.2%
Collected 181000 games, 224683 guess states, win rate: 88.2%
Collected 182000 games, 225922 guess states, win rate: 88.2%
Collected 183000 games, 227190 guess states, win rate: 88.2%
Collected 184000 games, 228432 guess states, win rate: 88.2%
Collected 185000 games, 229650 guess states, win rate: 88.2%
Collected 186000 games, 230893 guess states, win rate: 88.2%
Collected 187000 games, 232135 guess states, win rate: 88.2%
Collected 188000 games, 233388 guess states, win rate: 88.2%
Collected 189000 games, 234612 guess states, win rate: 88.2%
Collected 190000 games, 235854 guess states, win rate: 88.2%
Collected 191000 games, 237097 guess states, win rate: 88.2%
Collected 192000 games, 238362 guess states, win rate: 88.2%
Collected 193000 games, 239621 guess states, win rate: 88.2%
Collected 194000 games, 240885 guess states, win rate: 88.2%
Collected 195000 games, 242173 guess states, win rate: 88.2%
Collected 196000 games, 243407 guess states, win rate: 88.2%
Collected 197000 games, 244616 guess states, win rate: 88.2%
Collected 198000 games, 245844 guess states, win rate: 88.2%
Collected 199000 games, 247078 guess states, win rate: 88.2%
Collected 200000 games, 248361 guess states, win rate: 88.2%
Collected 248361 guess states in 679.0s
Phase 2: Training model for 50 epochs...
Epoch 1/50, Loss: 0.3690, LR: 0.001000
Epoch 2/50, Loss: 0.3663, LR: 0.001000
Epoch 3/50, Loss: 0.3659, LR: 0.001000
Epoch 4/50, Loss: 0.3657, LR: 0.001000
Epoch 5/50, Loss: 0.3654, LR: 0.001000
Epoch 6/50, Loss: 0.3652, LR: 0.001000
Epoch 7/50, Loss: 0.3649, LR: 0.001000
Epoch 8/50, Loss: 0.3645, LR: 0.001000
Epoch 9/50, Loss: 0.3642, LR: 0.001000
Epoch 10/50, Loss: 0.3638, LR: 0.000500
Epoch 11/50, Loss: 0.3628, LR: 0.000500
Epoch 12/50, Loss: 0.3623, LR: 0.000500
Epoch 13/50, Loss: 0.3618, LR: 0.000500
Epoch 14/50, Loss: 0.3615, LR: 0.000500
Epoch 15/50, Loss: 0.3611, LR: 0.000500
Epoch 16/50, Loss: 0.3608, LR: 0.000500
Epoch 17/50, Loss: 0.3604, LR: 0.000500
Epoch 18/50, Loss: 0.3601, LR: 0.000500
Epoch 19/50, Loss: 0.3598, LR: 0.000500
Epoch 20/50, Loss: 0.3596, LR: 0.000250
Epoch 21/50, Loss: 0.3586, LR: 0.000250
Epoch 22/50, Loss: 0.3583, LR: 0.000250
Epoch 23/50, Loss: 0.3580, LR: 0.000250
Epoch 24/50, Loss: 0.3578, LR: 0.000250
Epoch 25/50, Loss: 0.3576, LR: 0.000250
Epoch 26/50, Loss: 0.3575, LR: 0.000250
Epoch 27/50, Loss: 0.3573, LR: 0.000250
Epoch 28/50, Loss: 0.3571, LR: 0.000250
Epoch 29/50, Loss: 0.3570, LR: 0.000250
Epoch 30/50, Loss: 0.3568, LR: 0.000125
Epoch 31/50, Loss: 0.3562, LR: 0.000125
Epoch 32/50, Loss: 0.3560, LR: 0.000125
Epoch 33/50, Loss: 0.3559, LR: 0.000125
Epoch 34/50, Loss: 0.3558, LR: 0.000125
Epoch 35/50, Loss: 0.3557, LR: 0.000125
Epoch 36/50, Loss: 0.3556, LR: 0.000125
Epoch 37/50, Loss: 0.3555, LR: 0.000125
Epoch 38/50, Loss: 0.3555, LR: 0.000125
Epoch 39/50, Loss: 0.3554, LR: 0.000125
Epoch 40/50, Loss: 0.3553, LR: 0.000063
Epoch 41/50, Loss: 0.3549, LR: 0.000063
Epoch 42/50, Loss: 0.3548, LR: 0.000063
Epoch 43/50, Loss: 0.3548, LR: 0.000063
Epoch 44/50, Loss: 0.3548, LR: 0.000063
Epoch 45/50, Loss: 0.3547, LR: 0.000063
Epoch 46/50, Loss: 0.3547, LR: 0.000063
Epoch 47/50, Loss: 0.3546, LR: 0.000063
Epoch 48/50, Loss: 0.3546, LR: 0.000063
Epoch 49/50, Loss: 0.3545, LR: 0.000063
Epoch 50/50, Loss: 0.3545, LR: 0.000031
Training completed in 368.0s
Phase 3: Evaluating on 500 games...
Win rate: 95.2%, Guess accuracy: 96.3%
Exported ONNX model to models/easy_prob.onnx
============================================================
intermediate (16x16, 40 mines)
============================================================
Phase 1: Collecting training data from 300000 games...
Collected 1000 games, 1710 guess states, win rate: 61.5%
Collected 2000 games, 3447 guess states, win rate: 61.2%
Collected 3000 games, 5180 guess states, win rate: 61.5%
Collected 4000 games, 6867 guess states, win rate: 61.6%
Collected 5000 games, 8555 guess states, win rate: 61.6%
Collected 6000 games, 10334 guess states, win rate: 61.7%
Collected 7000 games, 12029 guess states, win rate: 61.5%
Collected 8000 games, 13797 guess states, win rate: 61.4%
Collected 9000 games, 15555 guess states, win rate: 61.1%
Collected 10000 games, 17233 guess states, win rate: 61.4%
Collected 11000 games, 18986 guess states, win rate: 61.2%
Collected 12000 games, 20669 guess states, win rate: 61.2%
Collected 13000 games, 22354 guess states, win rate: 61.3%
Collected 14000 games, 24142 guess states, win rate: 61.3%
Collected 15000 games, 25871 guess states, win rate: 61.4%
Collected 16000 games, 27555 guess states, win rate: 61.5%
Collected 17000 games, 29295 guess states, win rate: 61.5%
Collected 18000 games, 31035 guess states, win rate: 61.4%
Collected 19000 games, 32755 guess states, win rate: 61.3%
Collected 20000 games, 34474 guess states, win rate: 61.4%
Collected 21000 games, 36176 guess states, win rate: 61.5%
Collected 22000 games, 37875 guess states, win rate: 61.5%
Collected 23000 games, 39620 guess states, win rate: 61.5%
Collected 24000 games, 41336 guess states, win rate: 61.4%
Collected 25000 games, 43079 guess states, win rate: 61.4%
Collected 26000 games, 44798 guess states, win rate: 61.4%
Collected 27000 games, 46556 guess states, win rate: 61.4%
Collected 28000 games, 48286 guess states, win rate: 61.5%
Collected 29000 games, 50101 guess states, win rate: 61.5%
Collected 30000 games, 51800 guess states, win rate: 61.4%
Collected 31000 games, 53514 guess states, win rate: 61.5%
Collected 32000 games, 55247 guess states, win rate: 61.4%
Collected 33000 games, 56980 guess states, win rate: 61.5%
Collected 34000 games, 58749 guess states, win rate: 61.4%
Collected 35000 games, 60531 guess states, win rate: 61.5%
Collected 36000 games, 62257 guess states, win rate: 61.6%
Collected 37000 games, 63934 guess states, win rate: 61.6%
Collected 38000 games, 65657 guess states, win rate: 61.5%
Collected 39000 games, 67347 guess states, win rate: 61.6%
Collected 40000 games, 69068 guess states, win rate: 61.6%
Collected 41000 games, 70836 guess states, win rate: 61.5%
Collected 42000 games, 72576 guess states, win rate: 61.5%
Collected 43000 games, 74273 guess states, win rate: 61.5%
Collected 44000 games, 76050 guess states, win rate: 61.5%
Collected 45000 games, 77797 guess states, win rate: 61.5%
Collected 46000 games, 79550 guess states, win rate: 61.4%
Collected 47000 games, 81313 guess states, win rate: 61.4%
Collected 48000 games, 83071 guess states, win rate: 61.4%
Collected 49000 games, 84774 guess states, win rate: 61.5%
Collected 50000 games, 86489 guess states, win rate: 61.5%
Collected 51000 games, 88262 guess states, win rate: 61.6%
Collected 52000 games, 89960 guess states, win rate: 61.6%
Collected 53000 games, 91705 guess states, win rate: 61.6%
Collected 54000 games, 93382 guess states, win rate: 61.6%
Collected 55000 games, 95038 guess states, win rate: 61.6%
Collected 56000 games, 96779 guess states, win rate: 61.7%
Collected 57000 games, 98513 guess states, win rate: 61.6%
Collected 58000 games, 100208 guess states, win rate: 61.6%
Collected 59000 games, 101969 guess states, win rate: 61.7%
Collected 60000 games, 103703 guess states, win rate: 61.7%
Collected 61000 games, 105415 guess states, win rate: 61.7%
Collected 62000 games, 107169 guess states, win rate: 61.7%
Collected 63000 games, 108859 guess states, win rate: 61.7%
Collected 64000 games, 110497 guess states, win rate: 61.7%
Collected 65000 games, 112240 guess states, win rate: 61.7%
Collected 66000 games, 114014 guess states, win rate: 61.7%
Collected 67000 games, 115745 guess states, win rate: 61.7%
Collected 68000 games, 117543 guess states, win rate: 61.7%
Collected 69000 games, 119264 guess states, win rate: 61.7%
Collected 70000 games, 120952 guess states, win rate: 61.7%
Collected 71000 games, 122715 guess states, win rate: 61.7%
Collected 72000 games, 124464 guess states, win rate: 61.7%
Collected 73000 games, 126170 guess states, win rate: 61.7%
Collected 74000 games, 127951 guess states, win rate: 61.7%
Collected 75000 games, 129713 guess states, win rate: 61.7%
Collected 76000 games, 131347 guess states, win rate: 61.7%
Collected 77000 games, 133054 guess states, win rate: 61.7%
Collected 78000 games, 134796 guess states, win rate: 61.7%
Collected 79000 games, 136553 guess states, win rate: 61.7%
Collected 80000 games, 138310 guess states, win rate: 61.7%
Collected 81000 games, 140086 guess states, win rate: 61.7%
Collected 82000 games, 141854 guess states, win rate: 61.7%
Collected 83000 games, 143606 guess states, win rate: 61.7%
Collected 84000 games, 145349 guess states, win rate: 61.7%
Collected 85000 games, 147029 guess states, win rate: 61.7%
Collected 86000 games, 148778 guess states, win rate: 61.7%
Collected 87000 games, 150507 guess states, win rate: 61.7%
Collected 88000 games, 152276 guess states, win rate: 61.7%
Collected 89000 games, 154033 guess states, win rate: 61.7%
Collected 90000 games, 155738 guess states, win rate: 61.7%
Collected 91000 games, 157505 guess states, win rate: 61.7%
Collected 92000 games, 159248 guess states, win rate: 61.7%
Collected 93000 games, 160943 guess states, win rate: 61.7%
Collected 94000 games, 162700 guess states, win rate: 61.7%
Collected 95000 games, 164450 guess states, win rate: 61.7%
Collected 96000 games, 166169 guess states, win rate: 61.7%
Collected 97000 games, 167948 guess states, win rate: 61.7%
Collected 98000 games, 169659 guess states, win rate: 61.7%
Collected 99000 games, 171404 guess states, win rate: 61.7%
Collected 100000 games, 173187 guess states, win rate: 61.7%
Collected 101000 games, 174930 guess states, win rate: 61.7%
Collected 102000 games, 176636 guess states, win rate: 61.7%
Collected 103000 games, 178313 guess states, win rate: 61.7%
Collected 104000 games, 180039 guess states, win rate: 61.7%
Collected 105000 games, 181740 guess states, win rate: 61.7%
Collected 106000 games, 183479 guess states, win rate: 61.7%
Collected 107000 games, 185256 guess states, win rate: 61.6%
Collected 108000 games, 186982 guess states, win rate: 61.6%
Collected 109000 games, 188696 guess states, win rate: 61.6%
Collected 110000 games, 190448 guess states, win rate: 61.6%
Collected 111000 games, 192206 guess states, win rate: 61.6%
Collected 112000 games, 193920 guess states, win rate: 61.6%
Collected 113000 games, 195664 guess states, win rate: 61.6%
Collected 114000 games, 197384 guess states, win rate: 61.6%
Collected 115000 games, 199090 guess states, win rate: 61.6%
Collected 116000 games, 200833 guess states, win rate: 61.6%
Collected 117000 games, 202554 guess states, win rate: 61.6%
Collected 118000 games, 204281 guess states, win rate: 61.6%
Collected 119000 games, 206030 guess states, win rate: 61.6%
Collected 120000 games, 207770 guess states, win rate: 61.6%
Collected 121000 games, 209533 guess states, win rate: 61.6%
Collected 122000 games, 211342 guess states, win rate: 61.6%
Collected 123000 games, 213117 guess states, win rate: 61.6%
Collected 124000 games, 214821 guess states, win rate: 61.6%
Collected 125000 games, 216523 guess states, win rate: 61.6%
Collected 126000 games, 218308 guess states, win rate: 61.6%
Collected 127000 games, 220086 guess states, win rate: 61.6%
Collected 128000 games, 221833 guess states, win rate: 61.6%
Collected 129000 games, 223547 guess states, win rate: 61.6%
Collected 130000 games, 225269 guess states, win rate: 61.6%
Collected 131000 games, 226989 guess states, win rate: 61.6%
Collected 132000 games, 228717 guess states, win rate: 61.6%
Collected 133000 games, 230473 guess states, win rate: 61.6%
Collected 134000 games, 232220 guess states, win rate: 61.6%
Collected 135000 games, 233994 guess states, win rate: 61.6%
Collected 136000 games, 235753 guess states, win rate: 61.6%
Collected 137000 games, 237506 guess states, win rate: 61.6%
Collected 138000 games, 239216 guess states, win rate: 61.6%
Collected 139000 games, 241051 guess states, win rate: 61.6%
Collected 140000 games, 242818 guess states, win rate: 61.6%
Collected 141000 games, 244582 guess states, win rate: 61.6%
Collected 142000 games, 246315 guess states, win rate: 61.6%
Collected 143000 games, 248051 guess states, win rate: 61.6%
Collected 144000 games, 249794 guess states, win rate: 61.6%
Collected 145000 games, 251492 guess states, win rate: 61.6%
Collected 146000 games, 253239 guess states, win rate: 61.6%
Collected 147000 games, 255033 guess states, win rate: 61.6%
Collected 148000 games, 256787 guess states, win rate: 61.6%
Collected 149000 games, 258472 guess states, win rate: 61.6%
Collected 150000 games, 260186 guess states, win rate: 61.6%
Collected 151000 games, 261933 guess states, win rate: 61.6%
Collected 152000 games, 263720 guess states, win rate: 61.6%
Collected 153000 games, 265553 guess states, win rate: 61.5%
Collected 154000 games, 267252 guess states, win rate: 61.5%
Collected 155000 games, 269006 guess states, win rate: 61.6%
Collected 156000 games, 270751 guess states, win rate: 61.6%
Collected 157000 games, 272528 guess states, win rate: 61.5%
Collected 158000 games, 274225 guess states, win rate: 61.6%
Collected 159000 games, 275920 guess states, win rate: 61.5%
Collected 160000 games, 277632 guess states, win rate: 61.6%
Collected 161000 games, 279408 guess states, win rate: 61.6%
Collected 162000 games, 281089 guess states, win rate: 61.6%
Collected 163000 games, 282871 guess states, win rate: 61.5%
Collected 164000 games, 284600 guess states, win rate: 61.5%
Collected 165000 games, 286352 guess states, win rate: 61.5%
Collected 166000 games, 288153 guess states, win rate: 61.5%
Collected 167000 games, 289870 guess states, win rate: 61.5%
Collected 168000 games, 291564 guess states, win rate: 61.5%
Collected 169000 games, 293336 guess states, win rate: 61.5%
Collected 170000 games, 295127 guess states, win rate: 61.5%
Collected 171000 games, 296880 guess states, win rate: 61.5%
Collected 172000 games, 298621 guess states, win rate: 61.5%
Collected 173000 games, 300305 guess states, win rate: 61.5%
Collected 174000 games, 302097 guess states, win rate: 61.5%
Collected 175000 games, 303888 guess states, win rate: 61.5%
Collected 176000 games, 305596 guess states, win rate: 61.5%
Collected 177000 games, 307326 guess states, win rate: 61.5%
Collected 178000 games, 309028 guess states, win rate: 61.5%
Collected 179000 games, 310758 guess states, win rate: 61.5%
Collected 180000 games, 312454 guess states, win rate: 61.5%
Collected 181000 games, 314179 guess states, win rate: 61.5%
Collected 182000 games, 315963 guess states, win rate: 61.5%
Collected 183000 games, 317744 guess states, win rate: 61.5%
Collected 184000 games, 319470 guess states, win rate: 61.5%
Collected 185000 games, 321187 guess states, win rate: 61.5%
Collected 186000 games, 322921 guess states, win rate: 61.5%
Collected 187000 games, 324634 guess states, win rate: 61.5%
Collected 188000 games, 326450 guess states, win rate: 61.5%
Collected 189000 games, 328230 guess states, win rate: 61.5%
Collected 190000 games, 329968 guess states, win rate: 61.5%
Collected 191000 games, 331690 guess states, win rate: 61.5%
Collected 192000 games, 333463 guess states, win rate: 61.5%
Collected 193000 games, 335206 guess states, win rate: 61.5%
Collected 194000 games, 336959 guess states, win rate: 61.5%
Collected 195000 games, 338667 guess states, win rate: 61.5%
Collected 196000 games, 340405 guess states, win rate: 61.5%
Collected 197000 games, 342189 guess states, win rate: 61.5%
Collected 198000 games, 343945 guess states, win rate: 61.5%
Collected 199000 games, 345690 guess states, win rate: 61.5%
Collected 200000 games, 347485 guess states, win rate: 61.5%
Collected 201000 games, 349237 guess states, win rate: 61.5%
Collected 202000 games, 350907 guess states, win rate: 61.5%
Collected 203000 games, 352648 guess states, win rate: 61.5%
Collected 204000 games, 354391 guess states, win rate: 61.5%
Collected 205000 games, 356130 guess states, win rate: 61.5%
Collected 206000 games, 357852 guess states, win rate: 61.5%
Collected 207000 games, 359481 guess states, win rate: 61.5%
Collected 208000 games, 361201 guess states, win rate: 61.5%
Collected 209000 games, 362956 guess states, win rate: 61.5%
Collected 210000 games, 364711 guess states, win rate: 61.5%
Collected 211000 games, 366452 guess states, win rate: 61.5%
Collected 212000 games, 368194 guess states, win rate: 61.4%
Collected 213000 games, 369905 guess states, win rate: 61.4%
Collected 214000 games, 371577 guess states, win rate: 61.5%
Collected 215000 games, 373262 guess states, win rate: 61.5%
Collected 216000 games, 375007 guess states, win rate: 61.5%
Collected 217000 games, 376711 guess states, win rate: 61.5%
Collected 218000 games, 378480 guess states, win rate: 61.5%
Collected 219000 games, 380307 guess states, win rate: 61.5%
Collected 220000 games, 381954 guess states, win rate: 61.5%
Collected 221000 games, 383609 guess states, win rate: 61.5%
Collected 222000 games, 385338 guess states, win rate: 61.5%
Collected 223000 games, 387097 guess states, win rate: 61.5%
Collected 224000 games, 388835 guess states, win rate: 61.5%
Collected 225000 games, 390591 guess states, win rate: 61.5%
Collected 226000 games, 392285 guess states, win rate: 61.5%
Collected 227000 games, 394056 guess states, win rate: 61.5%
Collected 228000 games, 395852 guess states, win rate: 61.5%
Collected 229000 games, 397570 guess states, win rate: 61.5%
Collected 230000 games, 399296 guess states, win rate: 61.5%
Collected 231000 games, 400930 guess states, win rate: 61.5%
Collected 232000 games, 402725 guess states, win rate: 61.4%
Collected 233000 games, 404420 guess states, win rate: 61.5%
Collected 234000 games, 406158 guess states, win rate: 61.4%
Collected 235000 games, 407877 guess states, win rate: 61.4%
Collected 236000 games, 409692 guess states, win rate: 61.4%
Collected 237000 games, 411459 guess states, win rate: 61.4%
Collected 238000 games, 413157 guess states, win rate: 61.4%
Collected 239000 games, 414853 guess states, win rate: 61.4%
Collected 240000 games, 416570 guess states, win rate: 61.5%
Collected 241000 games, 418305 guess states, win rate: 61.5%
Collected 242000 games, 420058 guess states, win rate: 61.5%
Collected 243000 games, 421844 guess states, win rate: 61.5%
Collected 244000 games, 423564 guess states, win rate: 61.4%
Collected 245000 games, 425268 guess states, win rate: 61.4%
Collected 246000 games, 426920 guess states, win rate: 61.5%
Collected 247000 games, 428664 guess states, win rate: 61.5%
Collected 248000 games, 430452 guess states, win rate: 61.5%
Collected 249000 games, 432209 guess states, win rate: 61.5%
Collected 250000 games, 433948 guess states, win rate: 61.5%
Collected 251000 games, 435723 guess states, win rate: 61.5%
Collected 252000 games, 437401 guess states, win rate: 61.5%
Collected 253000 games, 439147 guess states, win rate: 61.5%
Collected 254000 games, 440951 guess states, win rate: 61.5%
Collected 255000 games, 442721 guess states, win rate: 61.5%
Collected 256000 games, 444479 guess states, win rate: 61.4%
Collected 257000 games, 446185 guess states, win rate: 61.4%
Collected 258000 games, 447927 guess states, win rate: 61.4%
Collected 259000 games, 449637 guess states, win rate: 61.4%
Collected 260000 games, 451398 guess states, win rate: 61.4%
Collected 261000 games, 453084 guess states, win rate: 61.4%
Collected 262000 games, 454884 guess states, win rate: 61.4%
Collected 263000 games, 456625 guess states, win rate: 61.4%
Collected 264000 games, 458322 guess states, win rate: 61.4%
Collected 265000 games, 460076 guess states, win rate: 61.4%
Collected 266000 games, 461863 guess states, win rate: 61.4%
Collected 267000 games, 463596 guess states, win rate: 61.4%
Collected 268000 games, 465293 guess states, win rate: 61.4%
Collected 269000 games, 467054 guess states, win rate: 61.4%
Collected 270000 games, 468760 guess states, win rate: 61.4%
Collected 271000 games, 470490 guess states, win rate: 61.4%
Collected 272000 games, 472221 guess states, win rate: 61.4%
Collected 273000 games, 473924 guess states, win rate: 61.4%
Collected 274000 games, 475603 guess states, win rate: 61.4%
Collected 275000 games, 477341 guess states, win rate: 61.4%
Collected 276000 games, 479059 guess states, win rate: 61.4%
Collected 277000 games, 480813 guess states, win rate: 61.4%
Collected 278000 games, 482531 guess states, win rate: 61.4%
Collected 279000 games, 484255 guess states, win rate: 61.4%
Collected 280000 games, 485960 guess states, win rate: 61.4%
Collected 281000 games, 487687 guess states, win rate: 61.4%
Collected 282000 games, 489436 guess states, win rate: 61.4%
Collected 283000 games, 491220 guess states, win rate: 61.4%
Collected 284000 games, 492926 guess states, win rate: 61.4%
Collected 285000 games, 494627 guess states, win rate: 61.4%
Collected 286000 games, 496373 guess states, win rate: 61.4%
Collected 287000 games, 498045 guess states, win rate: 61.4%
Collected 288000 games, 499828 guess states, win rate: 61.4%
Collected 289000 games, 501577 guess states, win rate: 61.4%
Collected 290000 games, 503410 guess states, win rate: 61.4%
Collected 291000 games, 505124 guess states, win rate: 61.4%
Collected 292000 games, 506846 guess states, win rate: 61.4%
Collected 293000 games, 508600 guess states, win rate: 61.4%
Collected 294000 games, 510336 guess states, win rate: 61.4%
Collected 295000 games, 512110 guess states, win rate: 61.4%
Collected 296000 games, 513891 guess states, win rate: 61.4%
Collected 297000 games, 515657 guess states, win rate: 61.4%
Collected 298000 games, 517356 guess states, win rate: 61.4%
Collected 299000 games, 519105 guess states, win rate: 61.4%
Collected 300000 games, 520869 guess states, win rate: 61.4%
Collected 520869 guess states in 12310.4s
Phase 2: Training model for 50 epochs...
Epoch 1/50, Loss: 0.4135, LR: 0.001000
Epoch 2/50, Loss: 0.4120, LR: 0.001000
Epoch 3/50, Loss: 0.4117, LR: 0.001000
Epoch 4/50, Loss: 0.4115, LR: 0.001000
Epoch 5/50, Loss: 0.4113, LR: 0.001000
Epoch 6/50, Loss: 0.4111, LR: 0.001000
Epoch 7/50, Loss: 0.4108, LR: 0.001000
Epoch 8/50, Loss: 0.4106, LR: 0.001000
Epoch 9/50, Loss: 0.4103, LR: 0.001000
Epoch 10/50, Loss: 0.4101, LR: 0.000500
Epoch 11/50, Loss: 0.4094, LR: 0.000500
Epoch 12/50, Loss: 0.4091, LR: 0.000500
Epoch 13/50, Loss: 0.4089, LR: 0.000500
Epoch 14/50, Loss: 0.4088, LR: 0.000500
Epoch 15/50, Loss: 0.4086, LR: 0.000500
Epoch 16/50, Loss: 0.4085, LR: 0.000500
Epoch 17/50, Loss: 0.4083, LR: 0.000500
Epoch 18/50, Loss: 0.4082, LR: 0.000500
Epoch 19/50, Loss: 0.4081, LR: 0.000500
Epoch 20/50, Loss: 0.4079, LR: 0.000250
Epoch 21/50, Loss: 0.4075, LR: 0.000250
Epoch 22/50, Loss: 0.4073, LR: 0.000250
Epoch 23/50, Loss: 0.4072, LR: 0.000250
Epoch 24/50, Loss: 0.4072, LR: 0.000250
Epoch 25/50, Loss: 0.4071, LR: 0.000250
Epoch 26/50, Loss: 0.4070, LR: 0.000250
Epoch 27/50, Loss: 0.4070, LR: 0.000250
Epoch 28/50, Loss: 0.4069, LR: 0.000250
Epoch 29/50, Loss: 0.4069, LR: 0.000250
Epoch 30/50, Loss: 0.4068, LR: 0.000125
Epoch 31/50, Loss: 0.4065, LR: 0.000125
Epoch 32/50, Loss: 0.4065, LR: 0.000125
Epoch 33/50, Loss: 0.4064, LR: 0.000125
Epoch 34/50, Loss: 0.4064, LR: 0.000125
Epoch 35/50, Loss: 0.4064, LR: 0.000125
Epoch 36/50, Loss: 0.4063, LR: 0.000125
Epoch 37/50, Loss: 0.4063, LR: 0.000125
Epoch 38/50, Loss: 0.4063, LR: 0.000125
Epoch 39/50, Loss: 0.4062, LR: 0.000125
Epoch 40/50, Loss: 0.4062, LR: 0.000063
Epoch 41/50, Loss: 0.4061, LR: 0.000063
Epoch 42/50, Loss: 0.4060, LR: 0.000063
Epoch 43/50, Loss: 0.4060, LR: 0.000063
Epoch 44/50, Loss: 0.4060, LR: 0.000063
Epoch 45/50, Loss: 0.4060, LR: 0.000063
Epoch 46/50, Loss: 0.4060, LR: 0.000063
Epoch 47/50, Loss: 0.4060, LR: 0.000063
Epoch 48/50, Loss: 0.4059, LR: 0.000063
Epoch 49/50, Loss: 0.4059, LR: 0.000063
Epoch 50/50, Loss: 0.4059, LR: 0.000031
Training completed in 1594.3s
Phase 3: Evaluating on 500 games...
Win rate: 84.6%, Guess accuracy: 91.4%
Exported ONNX model to models/intermediate_prob.onnx
Expert
(kits) z5362216@k173:~/minesweeper $ python train_probability.py --games 150000 --epochs 50 --difficulty expert --out models
Device: cuda
============================================================
expert (16x30, 99 mines)
============================================================
Phase 1: Collecting training data from 150000 games...
Collected 1000 games, 3396 guess states, win rate: 7.1%
Collected 2000 games, 6795 guess states, win rate: 7.5%
Collected 3000 games, 10206 guess states, win rate: 6.9%
Collected 4000 games, 13610 guess states, win rate: 6.7%
Collected 5000 games, 17080 guess states, win rate: 6.5%
Collected 6000 games, 20397 guess states, win rate: 6.4%
Collected 7000 games, 23683 guess states, win rate: 6.4%
Collected 8000 games, 27077 guess states, win rate: 6.5%
Collected 9000 games, 30489 guess states, win rate: 6.7%
Collected 10000 games, 33828 guess states, win rate: 6.8%
Collected 11000 games, 37187 guess states, win rate: 6.8%
Collected 12000 games, 40534 guess states, win rate: 6.7%
Collected 13000 games, 43994 guess states, win rate: 6.5%
Collected 14000 games, 47313 guess states, win rate: 6.6%
Collected 15000 games, 50845 guess states, win rate: 6.5%
Collected 16000 games, 54300 guess states, win rate: 6.5%
Collected 17000 games, 57742 guess states, win rate: 6.6%
Collected 18000 games, 61193 guess states, win rate: 6.6%
Collected 19000 games, 64676 guess states, win rate: 6.6%
Collected 20000 games, 68035 guess states, win rate: 6.6%
Collected 21000 games, 71382 guess states, win rate: 6.6%
Collected 22000 games, 74761 guess states, win rate: 6.6%
Collected 23000 games, 78167 guess states, win rate: 6.6%
Collected 24000 games, 81638 guess states, win rate: 6.6%
Collected 25000 games, 85046 guess states, win rate: 6.6%
Collected 26000 games, 88370 guess states, win rate: 6.6%
Collected 27000 games, 91762 guess states, win rate: 6.6%
Collected 28000 games, 95195 guess states, win rate: 6.6%
Collected 29000 games, 98568 guess states, win rate: 6.7%
Collected 30000 games, 101984 guess states, win rate: 6.7%
Collected 31000 games, 105401 guess states, win rate: 6.6%
Collected 32000 games, 108851 guess states, win rate: 6.7%
Collected 33000 games, 112147 guess states, win rate: 6.6%
Collected 34000 games, 115541 guess states, win rate: 6.6%
Collected 35000 games, 118887 guess states, win rate: 6.6%
Collected 36000 games, 122216 guess states, win rate: 6.6%
Collected 37000 games, 125533 guess states, win rate: 6.6%
Collected 38000 games, 128959 guess states, win rate: 6.6%
Collected 39000 games, 132404 guess states, win rate: 6.6%
Collected 40000 games, 135754 guess states, win rate: 6.6%
Collected 41000 games, 139137 guess states, win rate: 6.6%
Collected 42000 games, 142494 guess states, win rate: 6.6%
Collected 43000 games, 145726 guess states, win rate: 6.6%
Collected 44000 games, 149079 guess states, win rate: 6.6%
Collected 45000 games, 152407 guess states, win rate: 6.6%
Collected 46000 games, 155729 guess states, win rate: 6.6%
Collected 47000 games, 159214 guess states, win rate: 6.6%
Collected 48000 games, 162579 guess states, win rate: 6.6%
Collected 49000 games, 166053 guess states, win rate: 6.6%
Collected 50000 games, 169479 guess states, win rate: 6.6%
Collected 51000 games, 172918 guess states, win rate: 6.6%
Collected 52000 games, 176329 guess states, win rate: 6.6%
Collected 53000 games, 179731 guess states, win rate: 6.6%
Collected 54000 games, 183176 guess states, win rate: 6.6%
Collected 55000 games, 186700 guess states, win rate: 6.6%
Collected 56000 games, 190212 guess states, win rate: 6.6%
Collected 57000 games, 193670 guess states, win rate: 6.6%
Collected 58000 games, 196952 guess states, win rate: 6.6%
Collected 59000 games, 200479 guess states, win rate: 6.6%
Collected 60000 games, 203843 guess states, win rate: 6.6%
Collected 61000 games, 207240 guess states, win rate: 6.6%
Collected 62000 games, 210653 guess states, win rate: 6.6%
Collected 63000 games, 213984 guess states, win rate: 6.6%
Collected 64000 games, 217235 guess states, win rate: 6.7%
Collected 65000 games, 220669 guess states, win rate: 6.7%
Collected 66000 games, 224007 guess states, win rate: 6.7%
Collected 67000 games, 227448 guess states, win rate: 6.7%
Collected 68000 games, 230939 guess states, win rate: 6.7%
Collected 69000 games, 234375 guess states, win rate: 6.7%
Collected 70000 games, 237783 guess states, win rate: 6.7%
Collected 71000 games, 241120 guess states, win rate: 6.7%
Collected 72000 games, 244444 guess states, win rate: 6.7%
Collected 73000 games, 247740 guess states, win rate: 6.7%
Collected 74000 games, 251161 guess states, win rate: 6.7%
Collected 75000 games, 254453 guess states, win rate: 6.7%
Collected 76000 games, 257806 guess states, win rate: 6.6%
Collected 77000 games, 261325 guess states, win rate: 6.6%
Collected 78000 games, 264573 guess states, win rate: 6.6%
Collected 79000 games, 268167 guess states, win rate: 6.6%
Collected 80000 games, 271583 guess states, win rate: 6.6%
Collected 81000 games, 275035 guess states, win rate: 6.6%
Collected 82000 games, 278363 guess states, win rate: 6.6%
Collected 83000 games, 281799 guess states, win rate: 6.7%
Collected 84000 games, 285062 guess states, win rate: 6.7%
Collected 85000 games, 288506 guess states, win rate: 6.7%
Collected 86000 games, 291879 guess states, win rate: 6.7%
Collected 87000 games, 295369 guess states, win rate: 6.6%
Collected 88000 games, 298784 guess states, win rate: 6.6%
Collected 89000 games, 302275 guess states, win rate: 6.7%
Collected 90000 games, 305608 guess states, win rate: 6.7%
Collected 91000 games, 309026 guess states, win rate: 6.7%
Collected 92000 games, 312366 guess states, win rate: 6.7%
Collected 93000 games, 315799 guess states, win rate: 6.7%
Collected 94000 games, 319259 guess states, win rate: 6.7%
Collected 95000 games, 322666 guess states, win rate: 6.7%
Collected 96000 games, 325997 guess states, win rate: 6.7%
Collected 97000 games, 329313 guess states, win rate: 6.7%
Collected 98000 games, 332635 guess states, win rate: 6.7%
Collected 99000 games, 336163 guess states, win rate: 6.7%
Collected 100000 games, 339562 guess states, win rate: 6.7%
Collected 101000 games, 342972 guess states, win rate: 6.7%
Collected 102000 games, 346356 guess states, win rate: 6.7%
Collected 103000 games, 349836 guess states, win rate: 6.7%
Collected 104000 games, 353377 guess states, win rate: 6.7%
Collected 105000 games, 356872 guess states, win rate: 6.7%
Collected 106000 games, 360276 guess states, win rate: 6.7%
Collected 107000 games, 363604 guess states, win rate: 6.7%
Collected 108000 games, 366931 guess states, win rate: 6.7%
Collected 109000 games, 370389 guess states, win rate: 6.7%
Collected 110000 games, 373866 guess states, win rate: 6.6%
Collected 111000 games, 377228 guess states, win rate: 6.7%
Collected 112000 games, 380654 guess states, win rate: 6.7%
Collected 113000 games, 383979 guess states, win rate: 6.7%
Collected 114000 games, 387479 guess states, win rate: 6.6%
Collected 115000 games, 390858 guess states, win rate: 6.6%
Collected 116000 games, 394227 guess states, win rate: 6.7%
Collected 117000 games, 397699 guess states, win rate: 6.7%
Collected 118000 games, 401085 guess states, win rate: 6.7%
Collected 119000 games, 404420 guess states, win rate: 6.7%
Collected 120000 games, 407774 guess states, win rate: 6.7%
Collected 121000 games, 411237 guess states, win rate: 6.7%
Collected 122000 games, 414678 guess states, win rate: 6.7%
Collected 123000 games, 418072 guess states, win rate: 6.7%
Collected 124000 games, 421478 guess states, win rate: 6.7%
Collected 125000 games, 424988 guess states, win rate: 6.7%
Collected 126000 games, 428512 guess states, win rate: 6.7%
Collected 127000 games, 431938 guess states, win rate: 6.7%
Collected 128000 games, 435406 guess states, win rate: 6.7%
Collected 129000 games, 438986 guess states, win rate: 6.7%
Collected 130000 games, 442290 guess states, win rate: 6.7%
Collected 131000 games, 445644 guess states, win rate: 6.7%
Collected 132000 games, 448969 guess states, win rate: 6.7%
Collected 133000 games, 452481 guess states, win rate: 6.7%
Collected 134000 games, 455833 guess states, win rate: 6.7%
Collected 135000 games, 459220 guess states, win rate: 6.7%
Collected 136000 games, 462455 guess states, win rate: 6.7%
Collected 137000 games, 465801 guess states, win rate: 6.7%
Collected 138000 games, 469304 guess states, win rate: 6.7%
Collected 139000 games, 472793 guess states, win rate: 6.7%
Collected 140000 games, 476186 guess states, win rate: 6.7%
Collected 141000 games, 479520 guess states, win rate: 6.7%
Collected 142000 games, 482934 guess states, win rate: 6.7%
Collected 143000 games, 486290 guess states, win rate: 6.7%
Collected 144000 games, 489685 guess states, win rate: 6.7%
Collected 145000 games, 493117 guess states, win rate: 6.7%
Collected 146000 games, 496650 guess states, win rate: 6.7%
Collected 147000 games, 499935 guess states, win rate: 6.7%
Collected 148000 games, 503371 guess states, win rate: 6.7%
Collected 149000 games, 506730 guess states, win rate: 6.7%
Collected 150000 games, 510190 guess states, win rate: 6.7%
Collected 510190 guess states in 9818.7s
Phase 2: Training model for 50 epochs...
Epoch 1/50, Loss: 0.4736, LR: 0.001000
Epoch 2/50, Loss: 0.4720, LR: 0.001000
Epoch 3/50, Loss: 0.4716, LR: 0.001000
Epoch 4/50, Loss: 0.4713, LR: 0.001000
Epoch 5/50, Loss: 0.4710, LR: 0.001000
Epoch 6/50, Loss: 0.4707, LR: 0.001000
Epoch 7/50, Loss: 0.4704, LR: 0.001000
Epoch 8/50, Loss: 0.4701, LR: 0.001000
Epoch 9/50, Loss: 0.4698, LR: 0.001000
Epoch 10/50, Loss: 0.4695, LR: 0.000500
Epoch 11/50, Loss: 0.4688, LR: 0.000500
Epoch 12/50, Loss: 0.4686, LR: 0.000500
Epoch 13/50, Loss: 0.4684, LR: 0.000500
Epoch 14/50, Loss: 0.4683, LR: 0.000500
Epoch 15/50, Loss: 0.4681, LR: 0.000500
Epoch 16/50, Loss: 0.4680, LR: 0.000500
Epoch 17/50, Loss: 0.4679, LR: 0.000500
Epoch 18/50, Loss: 0.4678, LR: 0.000500
Epoch 19/50, Loss: 0.4677, LR: 0.000500
Epoch 20/50, Loss: 0.4676, LR: 0.000250
Epoch 21/50, Loss: 0.4672, LR: 0.000250
Epoch 22/50, Loss: 0.4671, LR: 0.000250
Epoch 23/50, Loss: 0.4670, LR: 0.000250
Epoch 24/50, Loss: 0.4669, LR: 0.000250
Epoch 25/50, Loss: 0.4669, LR: 0.000250
Epoch 26/50, Loss: 0.4668, LR: 0.000250
Epoch 27/50, Loss: 0.4668, LR: 0.000250
Epoch 28/50, Loss: 0.4667, LR: 0.000250
Epoch 29/50, Loss: 0.4667, LR: 0.000250
Epoch 30/50, Loss: 0.4667, LR: 0.000125
Epoch 31/50, Loss: 0.4664, LR: 0.000125
Epoch 32/50, Loss: 0.4664, LR: 0.000125
Epoch 33/50, Loss: 0.4663, LR: 0.000125
Epoch 34/50, Loss: 0.4663, LR: 0.000125
Epoch 35/50, Loss: 0.4663, LR: 0.000125
Epoch 36/50, Loss: 0.4663, LR: 0.000125
Epoch 37/50, Loss: 0.4663, LR: 0.000125
Epoch 38/50, Loss: 0.4662, LR: 0.000125
Epoch 39/50, Loss: 0.4662, LR: 0.000125
Epoch 40/50, Loss: 0.4662, LR: 0.000063
Epoch 41/50, Loss: 0.4660, LR: 0.000063
Epoch 42/50, Loss: 0.4660, LR: 0.000063
Epoch 43/50, Loss: 0.4660, LR: 0.000063
Epoch 44/50, Loss: 0.4660, LR: 0.000063
Epoch 45/50, Loss: 0.4660, LR: 0.000063
Epoch 46/50, Loss: 0.4660, LR: 0.000063
Epoch 47/50, Loss: 0.4660, LR: 0.000063
Epoch 48/50, Loss: 0.4660, LR: 0.000063
Epoch 49/50, Loss: 0.4660, LR: 0.000063
Epoch 50/50, Loss: 0.4659, LR: 0.000031
Training completed in 753.8s
Phase 3: Evaluating on 500 games...
Win rate: 38.4%, Guess accuracy: 87.7%
Exported ONNX model to models/expert_prob.onnx
============================================================
TRAINING COMPLETE
============================================================
expert: 38.4% win rate
Code
todo includes when filesystem stabilises.