Updated device

app.py

@@ -20,7 +20,7 @@ enc = tiktoken.get_encoding('gpt2')
 def inference(input_text, num_return_sequences, max_length):
     input_tokens = torch.tensor(enc.encode(input_text), dtype=torch.long)
     input_tokens = input_tokens.unsqueeze(0).repeat(num_return_sequences, 1)
-    x = input_tokens.to('cuda')
+    x = input_tokens.to(device)
 
     while x.size(1) < max_length:
         # forward the model to get the logits