diff --git a/test/registered/models_e2e/test_dsa_glm52_nvfp4_tp_mtp.py b/test/registered/models_e2e/test_dsa_glm52_nvfp4_tp_mtp.py index 25348d100..210112a5a 100644 --- a/test/registered/models_e2e/test_dsa_glm52_nvfp4_tp_mtp.py +++ b/test/registered/models_e2e/test_dsa_glm52_nvfp4_tp_mtp.py @@ -21,7 +21,7 @@ class TestGLM52NVFP4TPMTP( model = "nvidia/GLM-5.2-NVFP4" tp_size = 4 mem_fraction_static = 0.8 - bs_1_speed_thres = 300 + bs_1_speed_thres = 250 extra_server_args = [ "--moe-runner-backend", "flashinfer_trtllm",