From 6808c6d57165f6d48b50e4aec3c8323275cfc258 Mon Sep 17 00:00:00 2001 From: Baizhou Zhang Date: Tue, 4 Aug 2026 15:36:41 -0700 Subject: [PATCH] [Tiny] Little enhancement of Kimi-K3 test (#33609) --- test/registered/models_e2e/test_kimi_k3_b300.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/test/registered/models_e2e/test_kimi_k3_b300.py b/test/registered/models_e2e/test_kimi_k3_b300.py index 9da3bd023..9a7a378c1 100644 --- a/test/registered/models_e2e/test_kimi_k3_b300.py +++ b/test/registered/models_e2e/test_kimi_k3_b300.py @@ -16,7 +16,7 @@ from sglang.test.test_utils import ( popen_launch_server, ) -register_cuda_ci(est_time=1800, stage="base-c", runner_config="8-gpu-b300") +register_cuda_ci(est_time=900, stage="base-c", runner_config="8-gpu-b300") MODEL_PATH = ( "/data/radixark/model-cache/hub/models--moonshotai--Kimi-K3/" @@ -66,6 +66,7 @@ class TestKimiK3B300LowLatency(GSM8KMixin, CustomTestCase): DSPARK_DRAFT_MODEL, "--speculative-dspark-block-size", "7", + "--enable-linear-replayssm-spec", ], ) @@ -94,7 +95,6 @@ class TestKimiK3B300Balanced(GSM8KMixin, CustomTestCase): "8", "--dcp-size", "8", - "--disable-custom-all-reduce", "--mem-fraction-static", "0.85", "--weight-loader-prefetch-checkpoints",