Update README - Run 20251104_125328

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 78.67
+      value: 78.87
 ---
 
 # David: Multi-Scale Feature Classifier
@@ -51,15 +51,15 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 78.67%
-- **Best Epoch**: 3
-- **Final Train Accuracy**: 85.69%
+- **Validation Accuracy**: 78.87%
+- **Best Epoch**: 4
+- **Final Train Accuracy**: 87.75%
 
 ### Per-Scale Performance
-- **Scale 512**: 77.12%
-- **Scale 4096**: 77.92%
-- **Scale 8192**: 77.60%
-- **Scale 16384**: 78.06%
+- **Scale 512**: 77.27%
+- **Scale 4096**: 78.43%
+- **Scale 8192**: 78.03%
+- **Scale 16384**: 78.34%
 
 
 ## Usage
@@ -80,8 +80,8 @@ AbstractPhil/gated-david/
 │       └── 20251104_125328/
 │           ├── MODEL_SUMMARY.txt     # 🎯 Human-readable performance summary
 │           ├── training_history.json # 📈 Epoch-by-epoch training curve
-│           ├── best_model_acc78.67.safetensors  # ⭐ Accuracy in filename!
-│           ├── best_model_acc78.67_metadata.json
+│           ├── best_model_acc78.87.safetensors  # ⭐ Accuracy in filename!
+│           ├── best_model_acc78.87_metadata.json
 │           ├── final_model.safetensors
 │           ├── checkpoint_epoch_X_accYY.YY.safetensors
 │           ├── david_config.json
@@ -103,7 +103,7 @@ from huggingface_hub import hf_hub_download
 # Specify model variant and run
 model_name = "david_clip_vit_b16_cantor_decoupled_massive"
 run_id = "20251104_125328"
-accuracy = "78.67"  # From MODELS_INDEX.json
+accuracy = "78.87"  # From MODELS_INDEX.json
 
 # Download config
 config_path = hf_hub_download(
@@ -209,4 +209,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-11-04 13:23:51*
+*Generated on 2025-11-04 13:31:39*