Update serve/builder.py

serve/builder.py

@@ -109,5 +109,6 @@ def load_pretrained_model(model_path, model_base, model_name, model_type, load_8
 
     if model.generation_config.pad_token_id is None:
         model.generation_config.pad_token_id = model.generation_config.eos_token_id
-
+        
+    model.to('cuda')
     return tokenizer, model, image_processor, context_len