Simplify eagle tests and TP sync in grammar backend (#4066)

2025-03-04 13:40:40 -08:00
parent 03b0364f76
commit 77a3954bf7
14 changed files with 122 additions and 126 deletions
--- a/test/srt/test_vertex_endpoint.py
+++ b/test/srt/test_vertex_endpoint.py
@@ -24,6 +24,7 @@ class TestVertexEndpoint(unittest.TestCase):
            cls.model,
            cls.base_url,
            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=["--cuda-graph-max-bs", 2],
        )

    @classmethod