From b573e07f3e35cac5bf5c53216c68fec674331bb0 Mon Sep 17 00:00:00 2001 From: Zhanrui Sun <184402041+ZhanruiSunCh@users.noreply.github.com> Date: Tue, 9 Sep 2025 11:38:34 +0800 Subject: [PATCH] [None][infra] Disable CU12 build to save build time (cost > 5 hours on SBSA) (#7633) Signed-off-by: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com> --- jenkins/Build.groovy | 5 +++-- jenkins/L0_Test.groovy | 8 ++++---- 2 files changed, 7 insertions(+), 6 deletions(-) diff --git a/jenkins/Build.groovy b/jenkins/Build.groovy index 1ae232223f..c5c4429880 100644 --- a/jenkins/Build.groovy +++ b/jenkins/Build.groovy @@ -572,8 +572,9 @@ def launchStages(pipeline, cpu_arch, enableFailFast, globalVars) buildConfigs = [ "Build TRT-LLM": [LLM_DOCKER_IMAGE] + prepareLLMBuild( pipeline, cpu_arch == AARCH64_TRIPLE ? CONFIG_LINUX_AARCH64 : CONFIG_LINUX_X86_64_VANILLA), - "Build TRT-LLM CUDA12": [LLM_DOCKER_IMAGE_CU12] + prepareLLMBuild( - pipeline, cpu_arch == AARCH64_TRIPLE ? CONFIG_LINUX_AARCH64_CU12 : CONFIG_LINUX_X86_64_VANILLA_CU12), + // Disable CUDA12 build for too slow to build (cost > 5 hours on SBSA) + // "Build TRT-LLM CUDA12": [LLM_DOCKER_IMAGE_CU12] + prepareLLMBuild( + // pipeline, cpu_arch == AARCH64_TRIPLE ? CONFIG_LINUX_AARCH64_CU12 : CONFIG_LINUX_X86_64_VANILLA_CU12), "Build TRT-LLM LLVM": [LLM_DOCKER_IMAGE] + prepareLLMBuild( pipeline, cpu_arch == AARCH64_TRIPLE ? CONFIG_LINUX_AARCH64_LLVM : CONFIG_LINUX_X86_64_LLVM), "Build TRT-LLM Pybind": [LLM_DOCKER_IMAGE] + prepareLLMBuild( diff --git a/jenkins/L0_Test.groovy b/jenkins/L0_Test.groovy index a2e3cdd664..1705a1f6c1 100644 --- a/jenkins/L0_Test.groovy +++ b/jenkins/L0_Test.groovy @@ -950,10 +950,10 @@ def createKubernetesPodConfig(image, type, arch = "amd64", gpuCount = 1, perfMod if (type == "gh200") { hostnameMatch = """ - key: "kubernetes.io/hostname" - operator: In + operator: NotIn values: - - "lego-cg1-qct-066.ipp3a2.colossus" - - "lego-cg1-qct-069.ipp3a2.colossus\"""" + - "lego-cg1-qct-070.ipp3a2.colossus" + - "lego-cg1-qct-079.ipp3a2.colossus\"""" } def podConfig = [ @@ -2015,7 +2015,7 @@ def launchTestJobs(pipeline, testFilter) "A100X-PyTorch-1": ["a100x", "l0_a100", 1, 1], "L40S-PyTorch-1": ["l40s", "l0_l40s", 1, 2], "L40S-PyTorch-2": ["l40s", "l0_l40s", 2, 2], - "H100_PCIe-PyTorch-CU12-1": ["h100-cr", "l0_h100", 1, 3], + "H100_PCIe-PyTorch-1": ["h100-cr", "l0_h100", 1, 3], "H100_PCIe-PyTorch-2": ["h100-cr", "l0_h100", 2, 3], "H100_PCIe-PyTorch-3": ["h100-cr", "l0_h100", 3, 3], "H100_PCIe-CPP-1": ["h100-cr", "l0_h100", 1, 2],