best_model.json

{
  "model_name": "David-fully_shared-weighted_sum",
  "run_id": "20251012_235237",
  "timestamp": "2025-10-13T00:11:02.365363",
  "best_val_acc": 64.73133333333334,
  "best_epoch": 3,
  "final_train_acc": 61.30111583163371,
  "final_train_loss": 4.185147669827233,
  "scale_accuracies": {
    "256": 64.34866666666667,
    "512": 64.522
  },
  "architecture": {
    "preset": "small_fast",
    "sharing_mode": "fully_shared",
    "fusion_mode": "weighted_sum",
    "scales": [
      256,
      512
    ],
    "feature_dim": 512,
    "num_classes": 1000,
    "use_belly": false,
    "belly_expand": 2.0
  },
  "training": {
    "dataset": "AbstractPhil/imagenet-clip-features-orderly",
    "model_variant": [
      "clip_vit_b16",
      "clip_vit_laion_b32",
      "clip_vit_b32"
    ],
    "num_epochs": 10,
    "batch_size": 1024,
    "learning_rate": 0.01,
    "rose_weight": "0.1\u21920.8",
    "cayley_loss": false,
    "optimizer": "AdamW",
    "scheduler": "cosine_restarts"
  },
  "files": {
    "weights_safetensors": "weights/David-fully_shared-weighted_sum/20251012_235237/best_model_acc64.73.safetensors",
    "weights_pytorch": "weights/David-fully_shared-weighted_sum/20251012_235237/best_model.pth",
    "config": "weights/David-fully_shared-weighted_sum/20251012_235237/david_config.json",
    "training_config": "weights/David-fully_shared-weighted_sum/20251012_235237/train_config.json",
    "tensorboard": "runs/David-fully_shared-weighted_sum/20251012_235237/"
  }
}