============================================================
Teacher-Mark Binary Classifier Training
============================================================
  Data dir:    outlook/data_correction
  Device:      cuda
  Epochs:      100
  Batch size:  32
  LR:          0.001
  Latent dim:  64
  MP steps:    3
  Checkpoint:  outlook/ckpt/correction.pt
  Log:         /public/home/dxw/Codes/afem/outlook/ckpt/correction_train_log.json

Loading dataset...
  Found 1888 files across 100 samples
  Train samples: 80  Val samples: 20
  Train: 1513 graphs, 5301714 nodes, 158309 positive (3.0%)
  Val:   375 graphs, 1300911 nodes, 38849 positive (3.0%)
  Loaded in 11.4s

Training...
  Epoch   0 | train=1.2141 val=1.1141 | auc=0.7793 topk=0.182 phys=0.130 lr=1.00e-03 | 5.0s | total=5.0s
  Epoch  10 | train=0.6835 val=0.7105 | auc=0.9103 topk=0.338 phys=0.130 lr=1.00e-03 | 3.4s | total=39.4s
  Epoch  20 | train=0.6046 val=0.6748 | auc=0.9220 topk=0.370 phys=0.130 lr=1.00e-03 | 3.4s | total=73.5s
  Epoch  30 | train=0.5223 val=0.6292 | auc=0.9316 topk=0.402 phys=0.130 lr=1.00e-03 | 3.4s | total=107.5s
  Epoch  40 | train=0.4965 val=0.6377 | auc=0.9323 topk=0.421 phys=0.130 lr=1.00e-03 | 3.4s | total=141.5s
  Epoch  50 | train=0.4445 val=0.6005 | auc=0.9408 topk=0.433 phys=0.130 lr=5.00e-04 | 3.4s | total=175.4s
  Epoch  60 | train=0.4331 val=0.6315 | auc=0.9386 topk=0.438 phys=0.130 lr=5.00e-04 | 3.4s | total=209.4s
  Epoch  70 | train=0.4120 val=0.6089 | auc=0.9432 topk=0.452 phys=0.130 lr=2.50e-04 | 3.4s | total=243.5s
  Epoch  80 | train=0.4015 val=0.6321 | auc=0.9424 topk=0.455 phys=0.130 lr=1.25e-04 | 3.4s | total=277.7s
  Epoch  90 | train=0.3970 val=0.6364 | auc=0.9422 topk=0.455 phys=0.130 lr=6.25e-05 | 3.4s | total=311.7s
  Epoch  99 | train=0.3922 val=0.6449 | auc=0.9419 topk=0.454 phys=0.130 lr=6.25e-05 | 3.4s | total=342.5s

Training completed in 343.1s

Checkpoint saved: outlook/ckpt/correction.pt
Training log saved: /public/home/dxw/Codes/afem/outlook/ckpt/correction_train_log.json

============================================================
Best epoch: 57
  val_loss:  0.5873
  val_auc:   0.9412
  val_topk:  0.440
  phys_topk: 0.130
  GNN beats physics: YES
============================================================

Done.