Update generate.py

generate.py

@@ -11,7 +11,7 @@ def generate(model, input_ids, generation_config, left_padding=None, **kwargs):
         if pad_token is None:
             raise ValueError("pad_token is not defined")
         batch_size = input_ids.shape[0]
-        pad_tensor = torch.full(size=(batch_size, left_padding), fill_value=pad_token)
+        pad_tensor = torch.full(size=(batch_size, left_padding), fill_value=pad_token).to(input_ids.device)
         input_ids = torch.cat((input_ids, pad_tensor), dim=1)
 
     # Simple greedy decoding loop