Update README - Run 20251012_210041

README.md

@@ -21,7 +21,7 @@ model-index:
       type: imagenet-1k
     metrics:
     - type: accuracy
-      value: 75.74
+      value: 66.43
 ---
 
 # David: Multi-Scale Feature Classifier
@@ -45,25 +45,22 @@ exist simultaneously in the same shared space with the correct checks and spacin
 
 ### Training Configuration
 - **Dataset**: AbstractPhil/imagenet-clip-features-orderly
-- **Model Variant**: ['clip_vit_b16', 'clip_vit_laion_b32']
+- **Model Variant**: ['clip_vit_b16', 'clip_vit_laion_b32', 'clip_vit_b32']
 - **Epochs**: 10
 - **Batch Size**: 1024
 - **Learning Rate**: 0.01
 - **Rose Loss Weight**: 0.2 → 0.8
-- **Cayley Loss**: True
+- **Cayley Loss**: False
 
 ## Performance
 
 ### Best Results
-- **Validation Accuracy**: 75.74%
-- **Best Epoch**: 9
-- **Final Train Accuracy**: 83.82%
+- **Validation Accuracy**: 66.43%
+- **Best Epoch**: 0
+- **Final Train Accuracy**: 56.12%
 
 ### Per-Scale Performance
-- **Scale 256**: 74.18%
-- **Scale 512**: 75.73%
-- **Scale 768**: 76.05%
-- **Scale 1024**: 76.08%
+- **Scale 256**: 66.43%
 
 
 ## Usage
@@ -81,18 +78,18 @@ AbstractPhil/david-shared-space/
 ├── best_model.json                   # Latest best model info
 ├── weights/
 │   └── david_balanced/
-│       └── 20251012_194945/
+│       └── 20251012_210041/
 │           ├── MODEL_SUMMARY.txt     # 🎯 Human-readable performance summary
 │           ├── training_history.json # 📈 Epoch-by-epoch training curve
-│           ├── best_model_acc75.74.safetensors  # ⭐ Accuracy in filename!
-│           ├── best_model_acc75.74_metadata.json
+│           ├── best_model_acc66.43.safetensors  # ⭐ Accuracy in filename!
+│           ├── best_model_acc66.43_metadata.json
 │           ├── final_model.safetensors
 │           ├── checkpoint_epoch_X_accYY.YY.safetensors
 │           ├── david_config.json
 │           └── train_config.json
 └── runs/
     └── david_balanced/
-        └── 20251012_194945/
+        └── 20251012_210041/
             └── events.out.tfevents.* # TensorBoard logs
 ```
 
@@ -106,8 +103,8 @@ from huggingface_hub import hf_hub_download
 
 # Specify model variant and run
 model_name = "david_balanced"
-run_id = "20251012_194945"
-accuracy = "75.74"  # From MODELS_INDEX.json
+run_id = "20251012_210041"
+accuracy = "66.43"  # From MODELS_INDEX.json
 
 # Download config
 config_path = hf_hub_download(
@@ -184,7 +181,7 @@ 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→0.8)
-- **Cayley Loss**: Geometric regularization (enabled)
+- **Cayley Loss**: Geometric regularization (disabled)
 
 ### Optimization
 - **Optimizer**: AdamW
@@ -201,7 +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_194945}
+  note = {Run ID: 20251012_210041}
 }
 ```
 
@@ -216,4 +213,4 @@ Special thanks to Claude (Anthropic) for debugging assistance.
 
 ---
 
-*Generated on 2025-10-12 20:21:47*
+*Generated on 2025-10-12 21:05:14*