Update handler.py

handler.py

@@ -70,10 +70,10 @@ class EndpointHandler:
         inputs = self.processor(images=images, texts=prompt, return_tensors="pt")
         inputs["pixel_values"] = inputs["pixel_values"].unsqueeze(0)
         inputs["image_sizes"] = inputs["image_sizes"].unsqueeze(0)
-        inputs = inputs.to("cuda").to(torch.bfloat16)
+        inputs.to("cuda").to(torch.bfloat16)
 
         generation_args = {
-            "max_new_tokens": data.get("max_tokens", data.get("max_new_tokens", 128)),
+            "max_new_tokens": data.get("max_new_tokens", data.get("max_tokens", 128)),
             "temperature": data.get("temperature", 0.7),
             "do_sample": False,
             "use_cache": True,