diff --git a/llms/mlx_lm/utils.py b/llms/mlx_lm/utils.py
index 5ee043a3..4d815810 100644
--- a/llms/mlx_lm/utils.py
+++ b/llms/mlx_lm/utils.py
@@ -345,9 +345,7 @@ def stream_generate(
     with wired_limit(model, [generation_stream]):
         detokenizer.reset()
         tic = time.perf_counter()
-        for n, (token, logprobs) in enumerate(
-            generate_step(prompt, model, **kwargs),
-        ):
+        for n, (token, logprobs) in enumerate(generate_step(prompt, model, **kwargs)):
             if n == 0:
                 prompt_time = time.perf_counter() - tic
                 prompt_tps = prompt.size / prompt_time