============================================================
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 1956 files across 100 samples
  Train samples: 80  Val samples: 20
  Train: 1568 graphs, 2258286 nodes, 66961 positive (3.0%)
  Val:   388 graphs, 549666 nodes, 16295 positive (3.0%)
  Loaded in 7.4s

Training...
  Epoch   0 | train=1.3072 val=1.2934 | auc=0.6305 topk=0.127 phys=0.102 lr=1.00e-03 | 3.3s | total=3.3s
  Epoch  10 | train=0.8706 val=0.8948 | auc=0.8608 topk=0.258 phys=0.102 lr=1.00e-03 | 1.8s | total=21.9s
  Epoch  20 | train=0.7636 val=0.8598 | auc=0.8790 topk=0.312 phys=0.102 lr=1.00e-03 | 1.9s | total=40.6s
  Epoch  30 | train=0.6756 val=0.8004 | auc=0.8966 topk=0.319 phys=0.102 lr=1.00e-03 | 1.9s | total=59.2s
  Epoch  40 | train=0.6571 val=0.7746 | auc=0.8983 topk=0.343 phys=0.102 lr=1.00e-03 | 1.9s | total=77.8s
  Epoch  50 | train=0.5787 val=0.8101 | auc=0.9032 topk=0.366 phys=0.102 lr=5.00e-04 | 1.9s | total=96.4s
  Epoch  60 | train=0.5457 val=0.8059 | auc=0.9075 topk=0.372 phys=0.102 lr=2.50e-04 | 1.9s | total=115.0s
  Epoch  70 | train=0.5333 val=0.8503 | auc=0.9035 topk=0.372 phys=0.102 lr=1.25e-04 | 1.9s | total=133.6s
  Epoch  80 | train=0.5197 val=0.8692 | auc=0.9021 topk=0.368 phys=0.102 lr=1.25e-04 | 1.8s | total=152.1s
  Epoch  90 | train=0.5113 val=0.8570 | auc=0.9063 topk=0.375 phys=0.102 lr=6.25e-05 | 1.9s | total=170.7s
  Epoch  99 | train=0.5058 val=0.8703 | auc=0.9036 topk=0.371 phys=0.102 lr=3.13e-05 | 1.9s | total=187.4s

Training completed in 187.9s

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

============================================================
Best epoch: 37
  val_loss:  0.7372
  val_auc:   0.9060
  val_topk:  0.343
  phys_topk: 0.102
  GNN beats physics: YES
============================================================

Done.