best_model.json

{
  "model_name": "David-decoupled-cantor_scale",
  "run_id": "20251104_154540",
  "timestamp": "2025-11-04T15:57:33.702682",
  "best_val_acc": 78.904,
  "best_epoch": 4,
  "final_train_acc": 86.63156325443911,
  "final_train_loss": 1.4550095543688029,
  "scale_accuracies": {
    "256": 74.616,
    "512": 77.184,
    "768": 77.982,
    "1024": 77.986,
    "2048": 77.906,
    "4096": 77.97
  },
  "architecture": {
    "preset": "clip_vit_b16_cantor_big_window",
    "sharing_mode": "decoupled",
    "fusion_mode": "cantor_scale",
    "scales": [
      256,
      512,
      768,
      1024,
      2048,
      4096
    ],
    "feature_dim": 512,
    "num_classes": 1000,
    "use_belly": true,
    "belly_expand": 2.0
  },
  "training": {
    "dataset": "AbstractPhil/imagenet-clip-features-orderly",
    "model_variant": "clip_vit_b16",
    "num_epochs": 5,
    "batch_size": 512,
    "learning_rate": 0.001,
    "rose_weight": "0.1\u21920.5",
    "cayley_loss": false,
    "optimizer": "AdamW",
    "scheduler": "cosine_restarts"
  },
  "files": {
    "weights_safetensors": "weights/David-decoupled-cantor_scale/20251104_154540/best_model_acc78.90.safetensors",
    "weights_pytorch": "weights/David-decoupled-cantor_scale/20251104_154540/best_model.pth",
    "config": "weights/David-decoupled-cantor_scale/20251104_154540/david_config.json",
    "training_config": "weights/David-decoupled-cantor_scale/20251104_154540/train_config.json",
    "tensorboard": "runs/David-decoupled-cantor_scale/20251104_154540/"
  }
}