Update README - Run 20251012_050214

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 82.32
+      value: 82.37
 ---
 
 # David: Multi-Scale Crystal Classifier
@@ -51,20 +51,21 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 82.32%
-- **Best Epoch**: 8
-- **Final Train Accuracy**: 89.83%
+- **Validation Accuracy**: 82.37%
+- **Best Epoch**: 9
+- **Final Train Accuracy**: 90.26%
 
 ### Per-Scale Performance
-- **Scale 256**: 82.32%
-- **Scale 512**: 82.71%
-- **Scale 768**: 82.65%
-- **Scale 1024**: 82.46%
-- **Scale 1280**: 82.40%
-- **Scale 1536**: 82.35%
-- **Scale 1792**: 82.27%
-- **Scale 2048**: 82.28%
-- **Scale 2304**: 81.79%
+- **Scale 256**: 82.37%
+- **Scale 512**: 82.73%
+- **Scale 768**: 82.70%
+- **Scale 1024**: 82.53%
+- **Scale 1280**: 82.53%
+- **Scale 1536**: 82.48%
+- **Scale 1792**: 82.60%
+- **Scale 2048**: 82.55%
+- **Scale 2304**: 82.60%
+- **Scale 2560**: 82.17%
 
 
 ## Usage
@@ -187,4 +188,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 05:23:39*
+*Generated on 2025-10-12 05:26:08*