fix capture_bs when speculative decoding enabled (#10730)

2025-09-22 15:43:12 +08:00
parent 9d7e82a0ab
commit 944f1ea0ec
1 changed files with 1 additions and 1 deletions
--- a/python/sglang/srt/model_executor/cuda_graph_runner.py
+++ b/python/sglang/srt/model_executor/cuda_graph_runner.py
@@ -179,7 +179,7 @@ def get_batch_sizes_to_capture(model_runner: ModelRunner):
            capture_bs = (
                list(range(1, 9))
                + list(range(10, 33, 2))
-                + list(range(40, 64, 8))
+                + list(range(40, 65, 8))
                + list(range(80, 161, 16))
            )