Fix bug in multigpu setups

modules/text_generation.py

@@ -194,7 +194,7 @@ def generate_reply(question, max_new_tokens, do_sample, temperature, top_p, typi
             yield formatted_outputs(reply, shared.model_name)
 
             if not shared.args.flexgen:
-                if output[-1] == n:
+                if int(output[-1]) == int(n):
                     break
                 input_ids = torch.reshape(output, (1, output.shape[0]))
             else: