Update app.py

app.py

@@ -7,10 +7,11 @@ from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStream
 client = InferenceClient("BenBranyon/zephyr-sumbot-all-songs")
 
 #Transformers Code
-model_id = "BenBranyon/zephyr-sumbot-all-songs"
-model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", load_in_4bit=True)
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-tokenizer.use_default_system_prompt = False
+if torch.cuda.is_available():
+    model_id = "BenBranyon/zephyr-sumbot-all-songs"
+    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", load_in_4bit=True)
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    tokenizer.use_default_system_prompt = False
 
 
 def respond(