diff --git a/jenkins/L0_Test.groovy b/jenkins/L0_Test.groovy index a6fef33a63..4841100348 100644 --- a/jenkins/L0_Test.groovy +++ b/jenkins/L0_Test.groovy @@ -1155,7 +1155,7 @@ def runLLMTestlistWithSbatch(pipeline, platform, testList, config=VANILLA_CONFIG export pytestCommand="$pytestCommand" export coverageConfigFile="$coverageConfigFile" export NVIDIA_IMEX_CHANNELS=\${NVIDIA_IMEX_CHANNELS:-0} - export NVIDIA_VISIBLE_DEVICES=\${NVIDIA_VISIBLE_DEVICES:-\$(seq -s, 0 \$((\$(nvidia-smi --query-gpu=count -i 0 --format=noheader)-1)))} + export NVIDIA_VISIBLE_DEVICES=\${NVIDIA_VISIBLE_DEVICES:-\$(seq -s, 0 \$((\$(nvidia-smi --query-gpu=count -i 0 --format=csv,noheader)-1)))} ${envExportStatements} echo "Env NVIDIA_IMEX_CHANNELS: \$NVIDIA_IMEX_CHANNELS" @@ -3249,10 +3249,12 @@ def launchTestJobs(pipeline, testFilter) fullSet = parallelJobs.keySet() x86SlurmTestConfigs = [ - "DGX_H100-2_GPUs-PyTorch-Others-1": ["dgx-h100-x2-oci", "l0_dgx_h100", 1, 1, 2], + "DGX_H100-2_GPUs-PyTorch-Others-1": ["dgx-h100-x2-oci", "l0_dgx_h100", 1, 2, 2], + "DGX_H100-2_GPUs-PyTorch-Others-2": ["dgx-h100-x2-oci", "l0_dgx_h100", 2, 2, 2], "DGX_H100-2_GPUs-PyTorch-GptOss-1": ["dgx-h100-x2-oci", "l0_dgx_h100", 1, 1, 2], "DGX_H100-2_GPUs-PyTorch-Ray-1": ["dgx-h100-x2-oci", "l0_dgx_h100", 1, 1, 2], - "DGX_H100-4_GPUs-PyTorch-DeepSeek-1": ["dgx-h100-x4-oci", "l0_dgx_h100", 1, 1, 4], + "DGX_H100-4_GPUs-PyTorch-DeepSeek-1": ["dgx-h100-x4-oci", "l0_dgx_h100", 1, 2, 4], + "DGX_H100-4_GPUs-PyTorch-DeepSeek-2": ["dgx-h100-x4-oci", "l0_dgx_h100", 2, 2, 4], "DGX_H100-4_GPUs-PyTorch-GptOss-1": ["dgx-h100-x4-oci", "l0_dgx_h100", 1, 1, 4], "DGX_H100-4_GPUs-PyTorch-Others-1": ["dgx-h100-x4-oci", "l0_dgx_h100", 1, 1, 4], "DGX_H100-4_GPUs-PyTorch-Ray-1": ["dgx-h100-x4-oci", "l0_dgx_h100", 1, 1, 4],