Update README - Run 20251012_145649

README.md

@@ -12,7 +12,7 @@ datasets:
 metrics:
 - accuracy
 model-index:
-- name: David-fully_shared-weighted_sum
+- name: David-hierarchical-progressive
   results:
   - task:
       type: image-classification
@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 71.73
+      value: 72.33
 ---
 
 # David: Multi-Scale Crystal Classifier
@@ -32,12 +32,12 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Model Details
 
 ### Architecture
-- **Preset**: small_fast
-- **Sharing Mode**: fully_shared
-- **Fusion Mode**: weighted_sum
-- **Scales**: [256, 512]
+- **Preset**: hierarchical_refinement
+- **Sharing Mode**: hierarchical
+- **Fusion Mode**: progressive
+- **Scales**: [256, 512, 768, 1024]
 - **Feature Dim**: 512
-- **Parameters**: 656,898
+- **Parameters**: 10,493,184
 
 ### Training Configuration
 - **Dataset**: AbstractPhil/imagenet-clip-features-orderly
@@ -51,13 +51,15 @@ as class prototypes with role-weighted similarity computation (Rose Loss).
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 71.73%
-- **Best Epoch**: 9
-- **Final Train Accuracy**: 73.18%
+- **Validation Accuracy**: 72.33%
+- **Best Epoch**: 0
+- **Final Train Accuracy**: 68.34%
 
 ### Per-Scale Performance
-- **Scale 256**: 71.26%
-- **Scale 512**: 71.69%
+- **Scale 256**: 65.92%
+- **Scale 512**: 70.01%
+- **Scale 768**: 71.31%
+- **Scale 1024**: 71.29%
 
 
 ## Usage
@@ -74,19 +76,19 @@ AbstractPhil/gated-david/
 ├── README.md                         # This file
 ├── best_model.json                   # Latest best model info
 ├── weights/
-│   └── david_small_fast/
-│       └── 20251012_141246/
+│   └── david_hierarchical/
+│       └── 20251012_145649/
 │           ├── MODEL_SUMMARY.txt     # 🎯 Human-readable performance summary
 │           ├── training_history.json # 📈 Epoch-by-epoch training curve
-│           ├── best_model_acc71.73.safetensors  # ⭐ Accuracy in filename!
-│           ├── best_model_acc71.73_metadata.json
+│           ├── best_model_acc72.33.safetensors  # ⭐ Accuracy in filename!
+│           ├── best_model_acc72.33_metadata.json
 │           ├── final_model.safetensors
 │           ├── checkpoint_epoch_X_accYY.YY.safetensors
 │           ├── david_config.json
 │           └── train_config.json
 └── runs/
-    └── david_small_fast/
-        └── 20251012_141246/
+    └── david_hierarchical/
+        └── 20251012_145649/
             └── events.out.tfevents.* # TensorBoard logs
 ```
 
@@ -99,9 +101,9 @@ from huggingface_hub import hf_hub_download
 # Browse available models in MODELS_INDEX.json first!
 
 # Specify model variant and run
-model_name = "david_small_fast"
-run_id = "20251012_141246"
-accuracy = "71.73"  # From MODELS_INDEX.json
+model_name = "david_hierarchical"
+run_id = "20251012_145649"
+accuracy = "72.33"  # From MODELS_INDEX.json
 
 # Download config
 config_path = hf_hub_download(
@@ -150,7 +152,7 @@ with torch.no_grad():
 ## Architecture Overview
 
 ### Multi-Scale Processing
-David processes inputs at multiple scales (256, 512), 
+David processes inputs at multiple scales (256, 512, 768, 1024), 
 allowing it to capture both coarse and fine-grained features.
 
 ### Crystal Geometry
@@ -168,7 +170,7 @@ score = w_anchor * sim(z, anchor) + w_need * sim(z, need) + ...
 ```
 
 ### Fusion Strategy
-**weighted_sum**: Intelligently combines predictions from multiple scales.
+**progressive**: Intelligently combines predictions from multiple scales.
 
 ## Training Details
 
@@ -192,7 +194,7 @@ 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_141246}
+  note = {Run ID: 20251012_145649}
 }
 ```
 
@@ -207,4 +209,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 14:29:35*
+*Generated on 2025-10-12 14:58:46*