Add detailed logging for model loading debug

instruction_classifier.py

@@ -114,14 +114,21 @@ class InstructionClassifierSanitizer:
                 if hf_hub_download is None:
                     raise ImportError("huggingface_hub is not installed")
                 
+                print(f"🚀 Starting model download from {self.model_repo_id}")
+                print(f"   Device: {self.device}")
+                print(f"   Model name: {self.model_name}")
+                
                 # Use HF_TOKEN from environment for private repositories
                 token = os.getenv('HF_TOKEN')
                 if token:
                     print(f"📥 Downloading private model from {self.model_repo_id}...")
+                    print(f"   Using HF_TOKEN: {token[:8]}...{token[-8:] if len(token) > 16 else 'short'}")
                 else:
                     print(f"📥 Downloading public model from {self.model_repo_id}...")
+                    print("   No HF_TOKEN found - using public access")
                 
                 # Download the model file (returns file path, not model object)
+                print(f"   Downloading {self.model_filename}...")
                 model_path = hf_hub_download(
                     repo_id=self.model_repo_id,
                     filename=self.model_filename,
@@ -130,15 +137,30 @@ class InstructionClassifierSanitizer:
                 )
                 print(f"✅ Model file downloaded to: {model_path}")
                 
+                # Check file size
+                file_size = os.path.getsize(model_path) / (1024**3)  # GB
+                print(f"   File size: {file_size:.2f} GB")
+                
                 # Load the checkpoint from the downloaded file
+                print("🔄 Loading checkpoint into memory...")
                 checkpoint = torch.load(model_path, map_location=self.device)
+                print(f"   Checkpoint keys: {len(checkpoint.keys())}")
+                
                 self._load_model_weights(checkpoint)
                 print(f"✅ Model weights loaded from {self.model_repo_id}")
+                print(f"   Model parameter count: {sum(p.numel() for p in self.model.parameters())}")
+                
             except Exception as e:
-                print(f"❌ Failed to download model from {self.model_repo_id}: {e}")
-                print("Full error details:")
+                print(f"❌ CRITICAL ERROR: Failed to download model from {self.model_repo_id}")
+                print(f"   Error type: {type(e).__name__}")
+                print(f"   Error message: {e}")
+                print("   Full error details:")
                 import traceback
                 traceback.print_exc()
+                print("   Environment info:")
+                print(f"     HF_TOKEN set: {'Yes' if os.getenv('HF_TOKEN') else 'No'}")
+                print(f"     Device: {self.device}")
+                print(f"     PyTorch version: {torch.__version__}")
                 raise RuntimeError(f"Failed to download model from {self.model_repo_id}: {e}")
     
     def _load_model_weights(self, checkpoint):
@@ -461,9 +483,14 @@ def sanitize_tool_output(tool_output):
     Returns:
         Sanitized tool output with instruction content removed
     """
+    print(f"🔍 sanitize_tool_output called with: {tool_output[:100]}...")
+    
     sanitizer = get_sanitizer()
     if sanitizer is None:
         print("⚠️  Instruction classifier not available, returning original output")
         return tool_output
     
-    return sanitizer.sanitize_tool_output(tool_output)
+    print("✅ Sanitizer found, processing...")
+    result = sanitizer.sanitize_tool_output(tool_output)
+    print(f"🔒 Sanitization complete, result: {result[:100]}...")
+    return result