Update README - Run 20251012_050214

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 82.34
+      value: 79.48
 ---
 
 # David: Multi-Scale Crystal Classifier
@@ -32,17 +32,17 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Model Details
 
 ### Architecture
-- **Preset**: clip_vit_l14
+- **Preset**: clip_vit_l14_deep
 - **Sharing Mode**: partial_shared
 - **Fusion Mode**: deep_efficiency
-- **Scales**: [384, 768, 1024, 1280]
+- **Scales**: [256, 512, 768, 1024, 1280, 1536, 1792, 2048, 2304, 2560]
 - **Feature Dim**: 768
 - **Parameters**: ~8.8M
 
 ### Training Configuration
 - **Dataset**: AbstractPhil/imagenet-clip-features-orderly
 - **Model Variant**: clip_vit_l14
-- **Epochs**: 20
+- **Epochs**: 10
 - **Batch Size**: 1024
 - **Learning Rate**: 0.01
 - **Rose Loss Weight**: 0.1 → 0.5
@@ -51,15 +51,12 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 82.34%
-- **Best Epoch**: 7
-- **Final Train Accuracy**: 91.96%
+- **Validation Accuracy**: 79.48%
+- **Best Epoch**: 0
+- **Final Train Accuracy**: 67.24%
 
 ### Per-Scale Performance
-- **Scale 384**: 82.34%
-- **Scale 768**: 82.46%
-- **Scale 1024**: 82.43%
-- **Scale 1280**: 82.35%
+- **Scale 256**: 79.48%
 
 
 ## Usage
@@ -125,7 +122,7 @@ with torch.no_grad():
 ## Architecture Overview
 
 ### Multi-Scale Processing
-David processes inputs at multiple scales (384, 768, 1024, 1280), 
+David processes inputs at multiple scales (256, 512, 768, 1024, 1280, 1536, 1792, 2048, 2304, 2560), 
 allowing it to capture both coarse and fine-grained features.
 
 ### Crystal Geometry
@@ -167,7 +164,7 @@ score = w_anchor * sim(z, anchor) + w_need * sim(z, need) + ...
   author = {AbstractPhil},
   year = {2025},
   url = {https://huggingface.co/AbstractPhil/gated-david},
-  note = {Run ID: 20251012_041353}
+  note = {Run ID: 20251012_050214}
 }
 ```
 
@@ -182,4 +179,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 04:57:22*
+*Generated on 2025-10-12 05:04:37*