pytorch/.github/workflows/_binary-test-linux.yml
Huy Do 9035b6a651 Allow disable binary build jobs on CI (#100754)
Given the recent outage w.r.t. binary workflows running on CI, I want to close the gap between them and regular CI jobs.  The first part is to add the same filter step used by regular CI jobs so that oncalls can disable the job if need.

* Nightly runs are excluded as it includes the step to publish nightly binaries.  Allowing oncalls to disable this part requires more thoughts.  So this covers only CI binary build and test jobs
* As binary jobs doesn't have a concept of test matrix config which is a required parameter to the filter script, I use a pseudo input of test config default there

### Testing

* https://github.com/pytorch/pytorch/issues/100758.  The job is skipped in https://github.com/pytorch/pytorch/actions/runs/4911034089/jobs/8768782689
* https://github.com/pytorch/pytorch/issues/100759.  The job is skipped in https://github.com/pytorch/pytorch/actions/runs/4911033966/jobs/8768713669

Note that Windows binary jobs are not run in PR anymore after https://github.com/pytorch/pytorch/pull/100638, and MacOS binary jobs only run nightly.  So there are only Linux jobs left.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/100754
Approved by: https://github.com/ZainRizvi
2023-05-09 20:01:00 +00:00

212 lines
7.4 KiB
YAML

name: linux-binary-test
on:
workflow_call:
inputs:
build_name:
required: true
type: string
description: The build's name
build_environment:
required: true
type: string
description: The build environment
PYTORCH_ROOT:
required: true
type: string
description: Root directory for the pytorch/pytorch repository
BUILDER_ROOT:
required: true
type: string
description: Root directory for the pytorch/builder repository
PACKAGE_TYPE:
required: true
type: string
description: Package type
DESIRED_CUDA:
required: true
type: string
description: Desired Cuda version
GPU_ARCH_VERSION:
required: false
type: string
description: GPU Arch version
GPU_ARCH_TYPE:
required: true
type: string
description: GPU Arch type
DOCKER_IMAGE:
required: true
type: string
description: Docker image to use
LIBTORCH_CONFIG:
required: false
type: string
description: Desired libtorch config (for libtorch builds only)
LIBTORCH_VARIANT:
required: false
type: string
description: Desired libtorch variant (for libtorch builds only)
DESIRED_DEVTOOLSET:
required: false
type: string
description: Desired dev toolset
DESIRED_PYTHON:
required: false
type: string
description: Desired python version
runs_on:
required: true
type: string
description: Hardware to run this job on. Valid values are linux.4xlarge, linux.4xlarge.nvidia.gpu, and linux.rocm.gpu
secrets:
github-token:
required: true
description: Github Token
jobs:
test:
runs-on: ${{ inputs.runs_on }}
timeout-minutes: 240
env:
PYTORCH_ROOT: ${{ inputs.PYTORCH_ROOT }}
BUILDER_ROOT: ${{ inputs.BUILDER_ROOT }}
PACKAGE_TYPE: ${{ inputs.PACKAGE_TYPE }}
# TODO: This is a legacy variable that we eventually want to get rid of in
# favor of GPU_ARCH_VERSION
DESIRED_CUDA: ${{ inputs.DESIRED_CUDA }}
GPU_ARCH_VERSION: ${{ inputs.GPU_ARCH_VERSION }}
GPU_ARCH_TYPE: ${{ inputs.GPU_ARCH_TYPE }}
DOCKER_IMAGE: ${{ inputs.DOCKER_IMAGE }}
SKIP_ALL_TESTS: 1
LIBTORCH_CONFIG: ${{ inputs.LIBTORCH_CONFIG }}
LIBTORCH_VARIANT: ${{ inputs.LIBTORCH_VARIANT }}
DESIRED_DEVTOOLSET: ${{ inputs.DESIRED_DEVTOOLSET }}
DESIRED_PYTHON: ${{ inputs.DESIRED_PYTHON }}
# Needed for conda builds
ALPINE_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/tool/alpine"
ANACONDA_USER: pytorch
AWS_DEFAULT_REGION: us-east-1
BINARY_ENV_FILE: /tmp/env
BUILD_ENVIRONMENT: ${{ inputs.build_environment }}
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
PR_NUMBER: ${{ github.event.pull_request.number }}
PYTORCH_FINAL_PACKAGE_DIR: /artifacts
SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
steps:
- name: Make the env permanent during this workflow (but not the secrets)
shell: bash
run: |
{
echo "PYTORCH_ROOT=${{ env.PYTORCH_ROOT }}"
echo "BUILDER_ROOT=${{ env.BUILDER_ROOT }}"
echo "PACKAGE_TYPE=${{ env.PACKAGE_TYPE }}"
echo "DESIRED_CUDA=${{ env.DESIRED_CUDA }}"
echo "GPU_ARCH_VERSION=${{ env.GPU_ARCH_VERSION }}"
echo "GPU_ARCH_TYPE=${{ env.GPU_ARCH_TYPE }}"
echo "DOCKER_IMAGE=${{ env.DOCKER_IMAGE }}"
echo "SKIP_ALL_TESTS=${{ env.SKIP_ALL_TESTS }}"
echo "LIBTORCH_CONFIG=${{ env.LIBTORCH_CONFIG }}"
echo "LIBTORCH_VARIANT=${{ env.LIBTORCH_VARIANT }}"
echo "DESIRED_DEVTOOLSET=${{ env.DESIRED_DEVTOOLSET }}"
echo "DESIRED_PYTHON=${{ env.DESIRED_PYTHON }}"
echo "ALPINE_IMAGE=${{ env.ALPINE_IMAGE }}"
echo "ANACONDA_USER=${{ env.ANACONDA_USER }}"
echo "AWS_DEFAULT_REGION=${{ env.AWS_DEFAULT_REGION }}"
echo "BINARY_ENV_FILE=${{ env.BINARY_ENV_FILE }}"
echo "BUILD_ENVIRONMENT=${{ env.BUILD_ENVIRONMENT }}"
echo "PR_NUMBER=${{ env.PR_NUMBER }}"
echo "PYTORCH_FINAL_PACKAGE_DIR=${{ env.PYTORCH_FINAL_PACKAGE_DIR }}"
echo "SHA1=${{ env.SHA1 }}"
} >> "${GITHUB_ENV} }}"
- name: "[FB EMPLOYEES] Enable SSH (Click me for login details)"
uses: pytorch/test-infra/.github/actions/setup-ssh@main
with:
github-secret: ${{ secrets.github-token }}
# Setup the environment
- name: Checkout PyTorch
uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
- name: Setup Linux
uses: ./.github/actions/setup-linux
- name: Chown workspace
uses: ./.github/actions/chown-workspace
- name: Clean workspace
shell: bash
run: |
rm -rf "${GITHUB_WORKSPACE}"
mkdir "${GITHUB_WORKSPACE}"
- name: Checkout PyTorch to pytorch dir
uses: malfet/checkout@silent-checkout
with:
ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
submodules: recursive
path: pytorch
- name: Clean PyTorch checkout
run: |
# Remove any artifacts from the previous checkouts
git clean -fxd
working-directory: pytorch
- name: Checkout pytorch/builder to builder dir
uses: malfet/checkout@silent-checkout
with:
ref: main
submodules: recursive
repository: pytorch/builder
path: builder
- name: Clean pytorch/builder checkout
run: |
# Remove any artifacts from the previous checkouts
git clean -fxd
working-directory: builder
- name: Check if the job is disabled
id: filter
uses: ./pytorch/.github/actions/filter-test-configs
with:
github-token: ${{ secrets.GITHUB_TOKEN }}
# NB: Use a mock test matrix with a default value here. After filtering, if the
# returned matrix is empty, it means that the job is disabled
test-matrix: |
{ include: [
{ config: "default" },
]}
- name: Download Build Artifacts
if: ${{ steps.filter.outputs.is-test-matrix-empty == 'False' }}
uses: actions/download-artifact@v3
with:
name: ${{ inputs.build_name }}
path: "${{ runner.temp }}/artifacts/"
- name: Install nvidia driver, nvidia-docker runtime, set GPU_FLAG
uses: pytorch/test-infra/.github/actions/setup-nvidia@main
if: ${{ inputs.GPU_ARCH_TYPE == 'cuda' && steps.filter.outputs.is-test-matrix-empty == 'False' }}
- name: Pull Docker image
if: ${{ steps.filter.outputs.is-test-matrix-empty == 'False' }}
uses: pytorch/test-infra/.github/actions/pull-docker-image@main
with:
docker-image: ${{ inputs.DOCKER_IMAGE }}
- name: Test Pytorch binary
if: ${{ steps.filter.outputs.is-test-matrix-empty == 'False' }}
uses: ./pytorch/.github/actions/test-pytorch-binary
- name: Teardown Linux
if: always()
uses: pytorch/test-infra/.github/actions/teardown-linux@main
- name: Chown workspace
if: always()
uses: ./pytorch/.github/actions/chown-workspace