Update app.py

app.py

@@ -16,7 +16,8 @@ model = AutoModelForCausalLM.from_pretrained(
     repo_id, 
     token=access_token, 
     torch_dtype=torch.bfloat16,  # or use torch.bfloat16 if supported
-    device_map="auto"  # Automatically use available GPU/CPU efficiently
+    device_map="auto",  # Automatically use available GPU/CPU efficiently
+    low_cpu_mem_usage=True
 )
 
 def respond(