Update README - Run 20251012_235237

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 66.50
+      value: 61.64
 ---
 
 # David: Multi-Scale Feature Classifier
@@ -49,19 +49,18 @@ exist simultaneously in the same shared space with the correct checks and spacin
 - **Epochs**: 10
 - **Batch Size**: 1024
 - **Learning Rate**: 0.01
-- **Rose Loss Weight**: 0.2 → 0.6
+- **Rose Loss Weight**: 0.1 → 0.8
 - **Cayley Loss**: False
 
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 66.50%
-- **Best Epoch**: 7
-- **Final Train Accuracy**: 63.53%
+- **Validation Accuracy**: 61.64%
+- **Best Epoch**: 0
+- **Final Train Accuracy**: 54.39%
 
 ### Per-Scale Performance
-- **Scale 256**: 65.83%
-- **Scale 512**: 66.44%
+- **Scale 256**: 61.64%
 
 
 ## Usage
@@ -79,18 +78,18 @@ AbstractPhil/david-shared-space/
 ├── best_model.json                   # Latest best model info
 ├── weights/
 │   └── david_small_fast/
-│       └── 20251012_231445/
+│       └── 20251012_235237/
 │           ├── MODEL_SUMMARY.txt     # 🎯 Human-readable performance summary
 │           ├── training_history.json # 📈 Epoch-by-epoch training curve
-│           ├── best_model_acc66.50.safetensors  # ⭐ Accuracy in filename!
-│           ├── best_model_acc66.50_metadata.json
+│           ├── best_model_acc61.64.safetensors  # ⭐ Accuracy in filename!
+│           ├── best_model_acc61.64_metadata.json
 │           ├── final_model.safetensors
 │           ├── checkpoint_epoch_X_accYY.YY.safetensors
 │           ├── david_config.json
 │           └── train_config.json
 └── runs/
     └── david_small_fast/
-        └── 20251012_231445/
+        └── 20251012_235237/
             └── events.out.tfevents.* # TensorBoard logs
 ```
 
@@ -104,8 +103,8 @@ from huggingface_hub import hf_hub_download
 
 # Specify model variant and run
 model_name = "david_small_fast"
-run_id = "20251012_231445"
-accuracy = "66.50"  # From MODELS_INDEX.json
+run_id = "20251012_235237"
+accuracy = "61.64"  # From MODELS_INDEX.json
 
 # Download config
 config_path = hf_hub_download(
@@ -181,14 +180,14 @@ score = w_anchor * sim(z, anchor) + w_need * sim(z, need) + ...
 
 ### Loss Components
 - **Cross-Entropy**: Standard classification loss
-- **Rose Loss**: Pentachora role-weighted margin loss (weight: 0.2→0.6)
+- **Rose Loss**: Pentachora role-weighted margin loss (weight: 0.1→0.8)
 - **Cayley Loss**: Geometric regularization (disabled)
 
 ### Optimization
 - **Optimizer**: AdamW
 - **Weight Decay**: 1e-05
 - **Scheduler**: cosine_restarts
-- **Gradient Clip**: 5.0
+- **Gradient Clip**: 15.0
 - **Mixed Precision**: False
 
 ## Citation
@@ -199,7 +198,7 @@ score = w_anchor * sim(z, anchor) + w_need * sim(z, need) + ...
   author = {AbstractPhil},
   year = {2025},
   url = {https://huggingface.co/AbstractPhil/david-shared-space},
-  note = {Run ID: 20251012_231445}
+  note = {Run ID: 20251012_235237}
 }
 ```
 
@@ -214,4 +213,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 23:51:39*
+*Generated on 2025-10-12 23:57:17*