Use only one GPU for MLA CI tests (#2858)

2025-01-13 03:55:33 -08:00
parent 4536d72446
commit 67008f4b32
4 changed files with 39 additions and 8 deletions
--- a/test/srt/test_mla_fp8.py
+++ b/test/srt/test_mla_fp8.py
@@ -21,8 +21,6 @@ class TestMLA(unittest.TestCase):
            cls.base_url,
            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
            other_args=[
-                "--tp",
-                "2",
                "--trust-remote-code",
                "--kv-cache-dtype",
                "fp8_e5m2",