Update app.py

app.py

@@ -199,7 +199,14 @@ def run_inference(prompt, image, temperature=0.2, top_p=0.95, max_tokens=512):
             text_prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)
             inputs = processor(text=text_prompt, images=image, return_tensors="pt")
 
-        inputs = {k: v.to(model.device) for k, v in inputs.items()}
+        target_dtype = getattr(model, "dtype", None)
+        processed_inputs = {}
+        for key, value in inputs.items():
+            if torch.is_floating_point(value) and target_dtype is not None:
+                processed_inputs[key] = value.to(model.device, dtype=target_dtype)
+            else:
+                processed_inputs[key] = value.to(model.device)
+        inputs = processed_inputs
 
         with torch.no_grad():
             generated_ids = model.generate(
@@ -582,6 +589,9 @@ if api_key:
 # Load model at startup
 model_path = DEFAULT_MODEL
 logger.info(f"Loading model: {model_path}")
+
+
+
 # Check if it's a Qwen model
 if "qwenguard" in model_path.lower():
     model = Qwen2_5_VLForConditionalGeneration.from_pretrained(