From 6cca5b9b9795577f40aba0129dd26b6227d20c88 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=D0=90=D1=80=D1=82=D0=B5=D0=BC=20=D0=A1=D0=B0=D0=B2=D0=BA?= =?UTF-8?q?=D0=B8=D0=BD?= <58187114+OrangeRedeng@users.noreply.github.com> Date: Wed, 18 Mar 2026 22:47:43 +0300 Subject: [PATCH] [NPU] [BUGFIX] Test ascend memory consumption.py fix (#17995) --- test/registered/ascend/test_npu_memory_consumption.py | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/test/registered/ascend/test_npu_memory_consumption.py b/test/registered/ascend/test_npu_memory_consumption.py index 71fb7a02a..b229f8047 100644 --- a/test/registered/ascend/test_npu_memory_consumption.py +++ b/test/registered/ascend/test_npu_memory_consumption.py @@ -19,9 +19,8 @@ from sglang.test.test_utils import ( register_npu_ci( est_time=400, - suite="nightly-1-npu-a3", + suite="nightly-2-npu-a3", nightly=True, - disabled="run failed", ) if "ASCEND_RT_VISIBLE_DEVICES" not in os.environ: @@ -36,7 +35,7 @@ class TestMemoryConsumptionAscend(CustomTestCase): def test_memory_consumption(self): - model = "nytopop/Qwen3-30B-A3B.w8a8" + model = "/root/.cache/modelscope/hub/models/Qwen/Qwen3-30B-A3B-w8a8" base_url = DEFAULT_URL_FOR_TEST ### Calculate initial used memory @@ -71,7 +70,7 @@ class TestMemoryConsumptionAscend(CustomTestCase): used_memory_after_server_starting = ( total_npu_memory - free_npu_memory - initial_used_memory ) / (1 << 30) - self.assertLessEqual(float(used_memory_after_server_starting), 16.00) + self.assertLessEqual(float(used_memory_after_server_starting), 17.00) # Clean up everything kill_process_tree(process.pid)