[update]add main

main.py

@@ -42,11 +42,12 @@ def init_model(pretrained_model_name_or_path: str):
             offload_state_dict=True,
             # load_in_4bit=True,
         )
-        if model.config.model_type == "chatglm":
-            model = model.eval()
-        else:
-            model = model.to(device)
-            model = model.bfloat16().eval()
+        # if model.config.model_type == "chatglm":
+        #     model = model.eval()
+        # else:
+        #     model = model.to(device)
+        #     model = model.bfloat16().eval()
+        model = model.bfloat16().eval()
 
         tokenizer = AutoTokenizer.from_pretrained(
             pretrained_model_name_or_path,