onnxruntime/tools/ci_build/github/azure-pipelines/cuda-packaging-pipeline.yml
Yi Zhang d7aebf9ea8
Move Nuget Test from T4 to A10 to reduce release duration (#19253)
### Description
<!-- Describe your changes. -->



### Motivation and Context
Running release process is very painful and boring because some GPU jobs
have to wait so long time.

![image](https://github.com/microsoft/onnxruntime/assets/16190118/1c5c981e-68d4-4678-9758-443fbf362802)

![image](https://github.com/microsoft/onnxruntime/assets/16190118/ba0d79ba-1554-4c7a-93dd-6ea8144c9295)

![image](https://github.com/microsoft/onnxruntime/assets/16190118/36cab833-71c1-4ff5-bca5-f4caa9aee0c9)
On the one hand, we could move some T4 from PR process since some jobs
are not using T4 any more and on the other hand, we can continue to
change some jobs' agent from T4 to A4 too.

In the future, T4 will mainly be used for the scenarioes that big GPU
memory is needed, multiple GPU cards or some special cases.


Test runs:

https://dev.azure.com/aiinfra/Lotus/_build/results?buildId=401786&view=logs&j=8048494c-e6eb-5e47-5e87-ff0aa863325d

cc @YUNQIUGUO @snnn
2024-01-24 14:15:07 +08:00

197 lines
7 KiB
YAML

parameters:
- name: RunOnnxRuntimeTests
displayName: Run Tests?
type: boolean
default: true
- name: UseIncreasedTimeoutForTests
displayName: Increase timeout for tests? Set it to false if you are doing an Onnx Runtime release.
type: boolean
default: false
- name: DoCompliance
displayName: Run Compliance Tasks?
type: boolean
default: true
- name: DoEsrp
displayName: Run code sign tasks? Must be true if you are doing an ONNX Runtime release
type: boolean
default: true
- name: IsReleaseBuild
displayName: Is a release build? Set it to true if you are doing an ONNX Runtime release.
type: boolean
default: false
- name: PreReleaseVersionSuffixString
displayName: Suffix added to pre-release package version. Only used if IsReleaseBuild is true. Denotes the type of pre-release package.
type: string
values:
- alpha
- beta
- rc
- none
default: none
- name: PreReleaseVersionSuffixNumber
displayName: Number added to pre-release package version. Only used if IsReleaseBuild is true. Denotes the sequence of a pre-release package.
type: number
default: 0
# these 2 parameters are used for debugging.
- name: SpecificArtifact
displayName: Use Specific Artifact (Debugging only)
type: boolean
default: false
- name: BuildId
displayName: Pipeline BuildId, you could find it in the URL
type: string
default: '0'
- name: CudaVersion
displayName: CUDA version
type: string
default: '12.2'
values:
- 11.8
- 12.2
variables:
- name: ReleaseVersionSuffix
value: ''
- name: docker_base_image
${{ if eq(parameters.CudaVersion, '11.8') }}:
value: nvidia/cuda:11.8.0-cudnn8-devel-ubi8
${{ if eq(parameters.CudaVersion, '12.2') }}:
value: nvidia/cuda:12.2.2-cudnn8-devel-ubi8
- name: linux_trt_version
${{ if eq(parameters.CudaVersion, '11.8') }}:
value: 8.6.1.6-1.cuda11.8
${{ if eq(parameters.CudaVersion, '12.2') }}:
value: 8.6.1.6-1.cuda12.0
- name: win_trt_home
${{ if eq(parameters.CudaVersion, '11.8') }}:
value: $(Agent.TempDirectory)\TensorRT-8.6.1.6.Windows10.x86_64.cuda-11.8
${{ if eq(parameters.CudaVersion, '12.2') }}:
value: $(Agent.TempDirectory)\TensorRT-8.6.1.6.Windows10.x86_64.cuda-12.0
- name: win_cuda_home
${{ if eq(parameters.CudaVersion, '11.8') }}:
value: $(Agent.TempDirectory)\v11.8
${{ if eq(parameters.CudaVersion, '12.2') }}:
value: $(Agent.TempDirectory)\v12.2
resources:
repositories:
- repository: onnxruntime-inference-examples # The name used to reference this repository in the checkout step
type: github
endpoint: ort-examples
name: microsoft/onnxruntime-inference-examples
- repository: manylinux
type: Github
endpoint: Microsoft
name: pypa/manylinux
ref: 5eda9aded5462201e6310105728d33016e637ea7
stages:
# Set ReleaseVersionSuffix
- stage: Set_ReleaseVersionSuffix
jobs:
- job: Set_Variables
pool:
vmImage: ubuntu-latest
steps:
- checkout: none
- bash: |
# Do not output ##vso[] commands with `set -x` or they may be parsed again and include a trailing quote.
set +x
if [[ "${{ parameters.IsReleaseBuild }}" = True && "${{ parameters.PreReleaseVersionSuffixString }}" != "none" ]]; then
if [[ "${{ parameters.PreReleaseVersionSuffixNumber }}" -eq 0 ]]; then
echo "##vso[task.setvariable variable=ReleaseVersionSuffix;isOutput=true]-${{ parameters.PreReleaseVersionSuffixString }}"
else
echo "##vso[task.setvariable variable=ReleaseVersionSuffix;isOutput=true]-${{ parameters.PreReleaseVersionSuffixString }}.${{ parameters.PreReleaseVersionSuffixNumber }}"
fi
else
echo "##vso[task.setvariable variable=ReleaseVersionSuffix;isOutput=true]"
fi
name: Set_Release_Version_Suffix
- bash: echo $(ReleaseVersionSuffix)
name: Debug_Release_Version_Suffix
# this is needed for certain artifacts to be published
- stage: Linux_C_API_Packaging_CPU_x64
dependsOn: [ ]
jobs:
- template: templates/c-api-linux-cpu.yml
parameters:
BaseImage: 'registry.access.redhat.com/ubi8/ubi'
OnnxruntimeArch: 'x64'
OnnxruntimeNodejsBindingArch: 'x64'
PoolName: 'onnxruntime-Ubuntu2204-AMD-CPU'
PackageJava: false
PackageNodeJS: false
# Nuget Packaging
- template: stages/nuget-linux-cuda-packaging-stage.yml
parameters:
CudaVersion: ${{ parameters.CudaVersion }}
docker_base_image: ${{ variables.docker_base_image }}
linux_trt_version: ${{ variables.linux_trt_version }}
- template: stages/nuget-win-cuda-packaging-stage.yml
parameters:
RunOnnxRuntimeTests: ${{ parameters.RunOnnxRuntimeTests }}
UseIncreasedTimeoutForTests: ${{ parameters.UseIncreasedTimeoutForTests }}
CudaVersion: ${{ parameters.CudaVersion }}
win_trt_home: ${{ variables.win_trt_home }}
win_cuda_home: ${{ variables.win_cuda_home }}
- template: stages/nuget-combine-cuda-stage.yml
parameters:
DoCompliance: ${{ parameters.DoCompliance }}
DoEsrp: ${{ parameters.DoEsrp }}
IsReleaseBuild: ${{ parameters.IsReleaseBuild }}
# Testing
- template: nuget/templates/test_win.yml
parameters:
AgentPool : 'onnxruntime-Win2022-GPU-A10'
NugetPackageName : 'Microsoft.ML.OnnxRuntime.Gpu'
ArtifactSuffix: 'GPU'
StageSuffix: 'GPU'
Skipx86Tests: 'true'
CudaVersion: ${{ parameters.CudaVersion }}
SpecificArtifact: ${{ parameters.SpecificArtifact }}
BuildId: ${{ parameters.BuildId }}
- template: nuget/templates/test_win.yml
parameters:
AgentPool : 'onnxruntime-Win2022-GPU-A10'
NugetPackageName : 'Microsoft.ML.OnnxRuntime.Gpu.Windows'
ArtifactSuffix: 'GPU'
StageSuffix: 'GPU'
MoreSuffix: '_Windows'
Skipx86Tests: 'true'
CudaVersion: ${{ parameters.CudaVersion }}
SpecificArtifact: ${{ parameters.SpecificArtifact }}
BuildId: ${{ parameters.BuildId }}
- template: nuget/templates/test_linux.yml
parameters:
AgentPool : Onnxruntime-Linux-GPU-A10
ArtifactSuffix: 'GPU'
StageSuffix: 'GPU'
NugetPackageName : 'Microsoft.ML.OnnxRuntime.Gpu'
CudaVersion: ${{ parameters.CudaVersion }}
SpecificArtifact: ${{ parameters.specificArtifact }}
BuildId: ${{ parameters.BuildId }}
- template: nuget/templates/test_linux.yml
parameters:
AgentPool : Onnxruntime-Linux-GPU-A10
ArtifactSuffix: 'GPU'
StageSuffix: 'GPU'
MoreSuffix: '_Linux'
NugetPackageName : 'Microsoft.ML.OnnxRuntime.Gpu.Linux'
CudaVersion: ${{ parameters.CudaVersion }}
SpecificArtifact: ${{ parameters.specificArtifact }}
BuildId: ${{ parameters.BuildId }}
## Win/Linux GPU Combined Publishing
#- template: templates/publish-nuget.yml