Update app.py

app.py

@@ -45,7 +45,8 @@ class ModelManager:
             self.current_tokenizer = AutoTokenizer.from_pretrained(model_name)
             self.current_model = AutoModelForCausalLM.from_pretrained(
                 model_name, 
-                load_in_4bit=True,
+                load_in_4bit=False,
+                torch_dtype=torch.bfloat16,
                 device_map="auto"
             )
             self.current_model_name = model_name