| loaded 29717 examples from more_data/train.jsonl (0 skipped) |
| loaded 6226 examples from more_data/val.jsonl (0 skipped) |
| model: 230.32M params (230.32M active) on mps |
| config: {'tokenizer': 'tiktoken:o200k_base', 'vocab_size': 200029, 'd_model': 768, 'n_layers': 12, 'n_heads': 12, 'n_kv_heads': 4, 'd_ff': 2048, 'n_experts': 0, 'experts_top_k': 2, 'dropout': 0.0, 'rope_theta': 10000.0, 'readout_dim': 768, 'max_branch_tokens': 8192, 'max_total_tokens': 16384, 'max_options': 255, 'min_score_levels': 2, 'max_score_levels': 10} |
| epoch 0 step 10/450 decision=0.6495 lm=11.571 aux=0.000 lr=7.33e-05 19s mps_mem=16.1GB |
| epoch 0 step 20/450 decision=0.8738 lm=10.347 aux=0.000 lr=1.39e-04 33s mps_mem=15.7GB |
| epoch 0 step 30/450 decision=0.7129 lm=10.038 aux=0.000 lr=2.05e-04 49s mps_mem=16.3GB |
| epoch 0 step 40/450 decision=0.7050 lm=7.687 aux=0.000 lr=2.69e-04 66s mps_mem=16.0GB |
| epoch 0 step 50/450 decision=0.6913 lm=6.861 aux=0.000 lr=2.92e-04 81s mps_mem=16.3GB |
| epoch 0 step 60/450 decision=0.7137 lm=6.449 aux=0.000 lr=2.88e-04 98s mps_mem=16.3GB |
| epoch 0 step 70/450 decision=0.7306 lm=6.036 aux=0.000 lr=2.84e-04 111s mps_mem=16.4GB |
| epoch 0 step 80/450 decision=0.6081 lm=5.931 aux=0.000 lr=2.79e-04 126s mps_mem=16.4GB |
| epoch 0 step 90/450 decision=0.7980 lm=4.997 aux=0.000 lr=2.74e-04 139s mps_mem=16.4GB |
| epoch 0 step 100/450 decision=0.6342 lm=5.645 aux=0.000 lr=2.68e-04 154s mps_mem=16.6GB |
| epoch 0 step 110/450 decision=0.6476 lm=5.646 aux=0.000 lr=2.62e-04 169s mps_mem=17.6GB |
| epoch 0 step 120/450 decision=0.5984 lm=5.516 aux=0.000 lr=2.55e-04 185s mps_mem=16.0GB |
| epoch 0 step 130/450 decision=0.5570 lm=4.419 aux=0.000 lr=2.48e-04 200s mps_mem=16.4GB |
| epoch 0 step 140/450 decision=0.4154 lm=5.496 aux=0.000 lr=2.40e-04 215s mps_mem=15.9GB |
| epoch 0 step 150/450 decision=0.4894 lm=5.847 aux=0.000 lr=2.32e-04 229s mps_mem=16.9GB |
| epoch 0 step 160/450 decision=0.4100 lm=6.019 aux=0.000 lr=2.24e-04 245s mps_mem=17.6GB |
| epoch 0 step 170/450 decision=0.4492 lm=5.019 aux=0.000 lr=2.16e-04 261s mps_mem=17.4GB |
| epoch 0 step 180/450 decision=0.4949 lm=5.180 aux=0.000 lr=2.07e-04 279s mps_mem=17.2GB |
| epoch 0 step 190/450 decision=0.3374 lm=5.618 aux=0.000 lr=1.98e-04 294s mps_mem=17.6GB |
| epoch 0 step 200/450 decision=0.2344 lm=4.133 aux=0.000 lr=1.88e-04 308s mps_mem=19.1GB |
| epoch 0 step 210/450 decision=1.0210 lm=5.407 aux=0.000 lr=1.79e-04 323s mps_mem=18.2GB |
| epoch 0 step 220/450 decision=0.3333 lm=4.745 aux=0.000 lr=1.70e-04 336s mps_mem=18.4GB |
| [eval step 225] loss=0.5770 n_questions=11492 choice_acc=0.8788 choice_ece=0.0526 score_acc=0.9119 score_ece=0.0397 noul_acc=0.8310 noul_ece=0.0668 |
| epoch 0 step 230/450 decision=0.3640 lm=6.205 aux=0.000 lr=1.60e-04 399s mps_mem=19.1GB |
| epoch 0 step 240/450 decision=0.3701 lm=5.259 aux=0.000 lr=1.51e-04 414s mps_mem=19.0GB |
| epoch 0 step 250/450 decision=0.2392 lm=3.716 aux=0.000 lr=1.42e-04 432s mps_mem=18.7GB |
| epoch 0 step 260/450 decision=0.6256 lm=4.740 aux=0.000 lr=1.32e-04 446s mps_mem=18.7GB |
| epoch 0 step 270/450 decision=0.7182 lm=5.318 aux=0.000 lr=1.23e-04 460s mps_mem=19.5GB |
| epoch 0 step 280/450 decision=0.1141 lm=5.425 aux=0.000 lr=1.14e-04 475s mps_mem=18.8GB |
| epoch 0 step 290/450 decision=0.1184 lm=4.402 aux=0.000 lr=1.06e-04 491s mps_mem=17.9GB |
| epoch 0 step 300/450 decision=0.1633 lm=6.094 aux=0.000 lr=9.75e-05 506s mps_mem=18.2GB |
| epoch 0 step 310/450 decision=0.1656 lm=5.927 aux=0.000 lr=8.95e-05 520s mps_mem=18.3GB |
| epoch 0 step 320/450 decision=0.3844 lm=6.789 aux=0.000 lr=8.19e-05 535s mps_mem=19.3GB |
| epoch 0 step 330/450 decision=0.1993 lm=6.299 aux=0.000 lr=7.47e-05 549s mps_mem=18.3GB |
| epoch 0 step 340/450 decision=0.2600 lm=6.552 aux=0.000 lr=6.79e-05 565s mps_mem=19.0GB |
| epoch 0 step 350/450 decision=0.1330 lm=5.564 aux=0.000 lr=6.16e-05 582s mps_mem=18.2GB |
| epoch 0 step 360/450 decision=0.1488 lm=5.895 aux=0.000 lr=5.58e-05 598s mps_mem=18.6GB |
| epoch 0 step 370/450 decision=0.1584 lm=4.278 aux=0.000 lr=5.05e-05 613s mps_mem=17.7GB |
| epoch 0 step 380/450 decision=0.0673 lm=5.489 aux=0.000 lr=4.58e-05 628s mps_mem=18.7GB |
| epoch 0 step 390/450 decision=0.1097 lm=4.698 aux=0.000 lr=4.17e-05 642s mps_mem=18.2GB |
| epoch 0 step 400/450 decision=1.0655 lm=4.811 aux=0.000 lr=3.81e-05 658s mps_mem=18.3GB |
| epoch 0 step 410/450 decision=0.0921 lm=4.387 aux=0.000 lr=3.52e-05 672s mps_mem=17.8GB |
| epoch 0 step 420/450 decision=0.2600 lm=5.594 aux=0.000 lr=3.30e-05 685s mps_mem=18.7GB |
| epoch 0 step 430/450 decision=0.0631 lm=4.850 aux=0.000 lr=3.13e-05 702s mps_mem=18.6GB |
| epoch 0 step 440/450 decision=0.2136 lm=4.574 aux=0.000 lr=3.03e-05 720s mps_mem=17.8GB |
| epoch 0 step 450/450 decision=0.2333 lm=5.168 aux=0.000 lr=3.00e-05 736s mps_mem=18.8GB |
| [eval step 450] loss=0.3385 n_questions=11492 choice_acc=0.9213 choice_ece=0.0160 score_acc=0.9358 score_ece=0.0144 noul_acc=0.8479 noul_ece=0.0196 |
| [eval step 450] loss=0.3385 n_questions=11492 choice_acc=0.9213 choice_ece=0.0160 score_acc=0.9358 score_ece=0.0144 noul_acc=0.8479 noul_ece=0.0196 |
| temperatures: {'choice': 1.1402, 'score': 1.1319, 'noul': 1.0577} |
| [calibrated] loss=0.3343 n_questions=11492 choice_acc=0.9213 choice_ece=0.0184 score_acc=0.9358 score_ece=0.0096 noul_acc=0.8479 noul_ece=0.0191 |
| saved runs/medium-more/best.pt |
| |
