Update watchdog.py

watchdog.py

@@ -50,10 +50,10 @@ def fetch_training_data(tokenizer):
 
 def get_architecture_summary(model):
     summary = {
-        "Layers": getattr(model, "num_layers", "N/A"),
-        "Attention Heads": getattr(model, "num_heads", "N/A"),
-        "FFN Dim": getattr(model, "ffn_dim", "N/A"),
-        "Memory Enabled": getattr(model, "use_memory", "N/A"),
+        "Layers": getattr(model.config, "num_layers", "N/A"),
+        "Attention Heads": getattr(model.config, "num_heads", "N/A"),
+        "FFN Dim": getattr(model.config, "ffn_dim", "N/A"),
+        "Memory Enabled": getattr(model.config, "use_memory", "N/A"),
     }
     return "\n".join(f"{k}: {v}" for k, v in summary.items())
 
@@ -65,7 +65,16 @@ def retrain_model():
         if input_ids is None or len(input_ids) < 2:
             return "⚠️ Not enough data to retrain.", None, "Please log more feedback first."
 
-        config = EvoTransformerConfig()
+        # ✅ Explicitly define architecture details
+        config = EvoTransformerConfig(
+            hidden_size=384,
+            num_layers=6,
+            num_labels=2,
+            num_heads=6,
+            ffn_dim=1024,
+            use_memory=False
+        )
+
         model = EvoTransformerForClassification(config)
         model.train()
 
@@ -80,10 +89,10 @@ def retrain_model():
             optimizer.step()
             print(f"Epoch {epoch+1}: Loss = {loss.item():.4f}")
 
-        # Simulate accuracy (replace this with actual eval in the future)
-        accuracy = 1.0  # Replace with real evaluation later
+        # Simulate accuracy (placeholder)
+        accuracy = 1.0
 
-        # Save accuracy to evolution log
+        # Log evolution accuracy
         log_path = "trained_model/evolution_log.json"
         os.makedirs("trained_model", exist_ok=True)
 
@@ -98,11 +107,11 @@ def retrain_model():
         with open(log_path, "w") as f:
             json.dump(history, f)
 
-        # Save trained model
+        # Save model
         model.save_pretrained("trained_model")
         print("✅ EvoTransformer retrained and saved.")
 
-        # Reload the updated model for summary + plot
+        # Reload and return dashboard updates
         updated_model = load_model()
         arch_text = get_architecture_summary(updated_model)
         plot = evolution_accuracy_plot()
@@ -113,6 +122,6 @@ def retrain_model():
         print(f"❌ Retraining failed: {e}")
         return "❌ Error", None, f"Retrain failed: {e}"
 
-# Support direct script run
+# Allow direct script run
 if __name__ == "__main__":
     retrain_model()