Skip to content

feat: add aws_g6e instances #2571

feat: add aws_g6e instances

feat: add aws_g6e instances #2571

name: Components Integration Tests
on:
push:
branches:
- main
pull_request:
jobs:
components-launch:
strategy:
matrix:
include:
- scheduler: "aws_batch"
platform: "linux.20_04.4x"
- scheduler: "aws_batch"
container_repo: localhost
extra_args: "--mock"
platform: "linux.20_04.4x"
- scheduler: "kubernetes"
container_repo: localhost:5000/torchx
platform: "linux.20_04.16x"
- scheduler: "local_cwd"
platform: ubuntu-20.04
- scheduler: "local_docker"
platform: "linux.20_04.4x"
- scheduler: "ray"
platform: ubuntu-20.04
fail-fast: false
runs-on: ${{ matrix.platform }}
permissions:
id-token: write
contents: read
steps:
- name: Setup Python
uses: actions/setup-python@v2
with:
python-version: 3.8
architecture: x64
- name: Checkout TorchX
uses: actions/checkout@v2
- name: Configure AWS Credentials
uses: aws-actions/[email protected]
with:
aws-region: us-west-1
role-to-assume: ${{ secrets.TF_AWS_ROLE_ARN }}
role-session-name: github-torchx
continue-on-error: true
- name: Configure Docker
env:
AWS_ROLE_ARN: ${{ secrets.TF_AWS_ROLE_ARN }}
run: |
set -eux
if [ -n "$AWS_ROLE_ARN" ]; then
aws ecr get-login-password --region us-west-1 | docker login --username AWS --password-stdin 495572122715.dkr.ecr.us-west-1.amazonaws.com
fi
- name: Install dependencies
run: |
set -eux
pip install -r dev-requirements.txt
pip install -e .[kubernetes]
- name: Start Ray
if: ${{ matrix.scheduler == 'ray' }}
run: |
ray start --head
sleep 5 # wait till the node joins so that `status` command displays the head node info
ray status
- name: Start Kubernetes
if: ${{ matrix.scheduler == 'kubernetes' }}
run: |
scripts/setup_minikube.sh
- name: Run Components Integration Tests
env:
INTEGRATION_TEST_STORAGE: ${{ secrets.INTEGRATION_TEST_STORAGE }}
run: |
scripts/component_integration_tests.py --scheduler ${{ matrix.scheduler }} --container_repo "${{ matrix.container_repo || secrets.TF_CONTAINER_REPO }}" ${{ matrix.extra_args }}