Update app.py

app.py

@@ -13,7 +13,10 @@ auth_token = os.environ.get("AUTH_TOKEN_SECRET")
 tokenizer = LlamaTokenizer.from_pretrained("Claimed/capybara", use_auth_token=auth_token)
 
 model = LlamaForCausalLM.from_pretrained(
-     "Claimed/capybara", use_auth_token=auth_token, low_cpu_mem_usage=True)
+     "Claimed/capybara", use_auth_token=auth_token, 
+    load_in_8bit=True,
+    device_map="auto") #low_cpu_mem_usage=True)
+
 #model = model.to('cuda')