best_model.json

{
  "model_name": "David-partial_shared-hierarchical_tree",
  "run_id": "20251012_191456",
  "timestamp": "2025-10-12T19:24:50.899685",
  "best_val_acc": 71.268,
  "best_epoch": 2,
  "final_train_acc": 72.83585980594255,
  "final_train_loss": 2.280789690497776,
  "scale_accuracies": {
    "256": 70.197,
    "512": 70.734
  },
  "architecture": {
    "preset": "balanced",
    "sharing_mode": "partial_shared",
    "fusion_mode": "hierarchical_tree",
    "scales": [
      256,
      512,
      768,
      1024
    ],
    "feature_dim": 512,
    "num_classes": 1000,
    "use_belly": true,
    "belly_expand": 2.0
  },
  "training": {
    "dataset": "AbstractPhil/imagenet-clip-features-orderly",
    "model_variant": [
      "clip_vit_b32",
      "clip_vit_laion_b32"
    ],
    "num_epochs": 10,
    "batch_size": 1024,
    "learning_rate": 0.01,
    "rose_weight": "0.2\u21920.8",
    "cayley_loss": true,
    "optimizer": "AdamW",
    "scheduler": "cosine_restarts"
  },
  "files": {
    "weights_safetensors": "weights/David-partial_shared-hierarchical_tree/20251012_191456/best_model_acc71.27.safetensors",
    "weights_pytorch": "weights/David-partial_shared-hierarchical_tree/20251012_191456/best_model.pth",
    "config": "weights/David-partial_shared-hierarchical_tree/20251012_191456/david_config.json",
    "training_config": "weights/David-partial_shared-hierarchical_tree/20251012_191456/train_config.json",
    "tensorboard": "runs/David-partial_shared-hierarchical_tree/20251012_191456/"
  }
}