From 410088c91e4a6f9df5a071891578aa102262fe9b Mon Sep 17 00:00:00 2001 From: Cherry_ming <136634645@qq.com> Date: Mon, 10 Aug 2026 10:55:41 +0800 Subject: [PATCH] [NPU] Increase the retry count to 3 for the GSM8K. (#34103) Co-authored-by: Even Zhou Co-authored-by: sglang-npu-bot --- .github/workflows/_npu-single-node-test-stage.yml | 2 +- python/sglang/test/ascend/e2e/test_npu_accuracy_utils.py | 4 +--- 2 files changed, 2 insertions(+), 4 deletions(-) diff --git a/.github/workflows/_npu-single-node-test-stage.yml b/.github/workflows/_npu-single-node-test-stage.yml index f04757db7..be5bda0e4 100644 --- a/.github/workflows/_npu-single-node-test-stage.yml +++ b/.github/workflows/_npu-single-node-test-stage.yml @@ -260,7 +260,7 @@ jobs: echo "Running test: ${tc_name}" test_exit_code=0 cd test - ${PYTHON_FOR_SGLANG} -u run_suite.py --hw npu --suite ${test_suite} --timeout-per-file 3600 2>&1 | tee /tmp/test_output.log || test_exit_code=$? + ${PYTHON_FOR_SGLANG} -u run_suite.py --hw npu --suite ${test_suite} --timeout-per-file 7200 2>&1 | tee /tmp/test_output.log || test_exit_code=$? echo "Finished test: ${tc_name}" diff --git a/python/sglang/test/ascend/e2e/test_npu_accuracy_utils.py b/python/sglang/test/ascend/e2e/test_npu_accuracy_utils.py index 5fcdf3d5c..94d746c9d 100644 --- a/python/sglang/test/ascend/e2e/test_npu_accuracy_utils.py +++ b/python/sglang/test/ascend/e2e/test_npu_accuracy_utils.py @@ -61,6 +61,7 @@ DATASET_FLUCTUATION = { "aime25": 2, "aime26": 2, "gpqa_diamond": 5, + "gsm8k": 3, } MAX_RETRY_COUNT = 3 @@ -291,7 +292,6 @@ class TestNpuAccuracyTestCaseBase(CustomTestCase): other_args = None server_timeout = DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH envs = None - max_attempts = 2 n_runs = 3 accuracy = 0.1 test_type = "accuracy" @@ -549,7 +549,6 @@ class TestNpuAccuracyMultiNodePdMixTestCaseBase(CustomTestCase): other_args = None server_timeout = DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH envs = None - max_attempts = 2 accuracy = 0.1 @classmethod @@ -651,7 +650,6 @@ class TestNpuAccuracyMultiNodePdSepTestCaseBase(CustomTestCase): eval_type = "openai_api" other_args = None server_timeout = DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH - max_attempts = 2 accuracy = 0.1 @classmethod