Update README - Run 20251012_135249

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 69.62
+      value: 70.14
 ---
 
 # David: Multi-Scale Crystal Classifier
@@ -51,13 +51,13 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 69.62%
-- **Best Epoch**: 1
-- **Final Train Accuracy**: 69.48%
+- **Validation Accuracy**: 70.14%
+- **Best Epoch**: 2
+- **Final Train Accuracy**: 70.34%
 
 ### Per-Scale Performance
-- **Scale 256**: 68.83%
-- **Scale 512**: 69.57%
+- **Scale 256**: 69.54%
+- **Scale 512**: 70.03%
 
 
 ## Usage
@@ -78,8 +78,8 @@ AbstractPhil/gated-david/
 │       └── 20251012_135249/
 │           ├── MODEL_SUMMARY.txt     # 🎯 Human-readable performance summary
 │           ├── training_history.json # 📈 Epoch-by-epoch training curve
-│           ├── best_model_acc69.62.safetensors  # ⭐ Accuracy in filename!
-│           ├── best_model_acc69.62_metadata.json
+│           ├── best_model_acc70.14.safetensors  # ⭐ Accuracy in filename!
+│           ├── best_model_acc70.14_metadata.json
 │           ├── final_model.safetensors
 │           ├── checkpoint_epoch_X_accYY.YY.safetensors
 │           ├── david_config.json
@@ -101,7 +101,7 @@ from huggingface_hub import hf_hub_download
 # Specify model variant and run
 model_name = "david_small_fast"
 run_id = "20251012_135249"
-accuracy = "69.62"  # From MODELS_INDEX.json
+accuracy = "70.14"  # From MODELS_INDEX.json
 
 # Download config
 config_path = hf_hub_download(
@@ -207,4 +207,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 13:56:15*
+*Generated on 2025-10-12 13:57:57*