Files
pytorch/.github/workflows/_rocm-test.yml
Jithun Nair 0ec0120b19 Move aws OIDC credentials steps into setup-rocm.yml (#164769)
The AWS ECR login step needs `id-token: write` permissions. We move the steps to get OIDC-based credentials from `_rocm-test.yml` to `setup-rocm.yml`. This lays the groundwork to enable access to AWS ECR in workflows in other repos such as torchtitan that use [linux_job_v2.yml](https://github.com/pytorch/test-infra/blob/main/.github/workflows/linux_job_v2.yml), which also uses [setup-rocm.yml](335f4f80a0/.github/workflows/linux_job_v2.yml (L168)).

Any caller workflows that eventually execute `setup-rocm` action will thus need to provide the `id-token: write` permission.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/164769
Approved by: https://github.com/huydhn
2025-10-10 21:24:29 +00:00

355 lines
14 KiB
YAML

# TODO: this looks sort of similar to _linux-test, but there are like a dozen
# places where you would have to insert an if statement. Probably it's better to
# just use a different workflow altogether
name: test
on:
workflow_call:
inputs:
build-environment:
required: true
type: string
description: Top-level label for what's being built/tested.
test-matrix:
required: true
type: string
description: JSON description of what test configs to run.
docker-image:
required: true
type: string
description: Docker image to run in.
sync-tag:
required: false
type: string
default: ""
description: |
If this is set, our linter will use this to make sure that every other
job with the same `sync-tag` is identical.
timeout-minutes:
required: false
type: number
default: 300
description: |
Set the maximum (in minutes) how long the workflow should take to finish
tests-to-include:
required: false
type: string
default: ""
description: |
List of tests to include (empty string implies default list)
dashboard-tag:
required: false
type: string
default: ""
disable-monitor:
description: |
[Experimental] Disable utilization monitoring for tests.
Currently, by default we disable the monitor job and only look for specific tests,
since we are investigating the behaviour of the monitor script with different tests.
required: false
type: boolean
default: true
monitor-log-interval:
description: |
Set the interval for the monitor script to log utilization.
required: false
type: number
default: 5
monitor-data-collect-interval:
description: |
Set the interval for the monitor script to collect data.
required: false
type: number
default: 1
secrets:
HUGGING_FACE_HUB_TOKEN:
required: false
description: |
HF Auth token to avoid rate limits when downloading models or datasets from hub
env:
GIT_DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
permissions:
id-token: write
contents: read
jobs:
test:
# Don't run on forked repos or empty test matrix
if: github.repository_owner == 'pytorch' && toJSON(fromJSON(inputs.test-matrix).include) != '[]'
strategy:
matrix: ${{ fromJSON(inputs.test-matrix) }}
fail-fast: false
runs-on: ${{ matrix.runner }}
timeout-minutes: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
steps:
- name: Checkout PyTorch
uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
with:
no-sudo: true
- name: Setup ROCm
uses: ./.github/actions/setup-rocm
- name: Runner check GPU count (distributed jobs)
if: ${{ contains(matrix.config, 'distributed') }}
shell: bash
run: |
ngpu=$(rocminfo | grep -c -E 'Name:.*\sgfx')
if [[ $ngpu -lt 4 ]]; then
echo "Error: only $ngpu GPU(s) detected, at least 4 GPUs are needed for distributed jobs"
exit 1
fi
- name: Calculate docker image
id: calculate-docker-image
uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
with:
docker-image-name: ${{ inputs.docker-image }}
- name: Pull docker image
uses: pytorch/test-infra/.github/actions/pull-docker-image@main
with:
docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
- name: Get workflow job id
id: get-job-id
uses: ./.github/actions/get-workflow-job-id
if: always()
with:
github-token: ${{ secrets.GITHUB_TOKEN }}
- name: Start monitoring script
id: monitor-script
if: ${{ !inputs.disable-monitor }}
shell: bash
continue-on-error: true
env:
JOB_ID: ${{ steps.get-job-id.outputs.job-id }}
JOB_NAME: ${{ steps.get-job-id.outputs.job-name }}
WORKFLOW_NAME: ${{ github.workflow }}
WORKFLOW_RUN_ID: ${{github.run_id}}
MONITOR_LOG_INTERVAL: ${{ inputs.monitor-log-interval }}
MONITOR_DATA_COLLECT_INTERVAL: ${{ inputs.monitor-data-collect-interval }}
run: |
python3 -m pip install psutil==5.9.8 dataclasses_json==0.6.7
python3 -m tools.stats.monitor --log-interval "$MONITOR_LOG_INTERVAL" --data-collect-interval "$MONITOR_DATA_COLLECT_INTERVAL" > usage_log.txt 2>&1 &
echo "monitor-script-pid=${!}" >> "${GITHUB_OUTPUT}"
- name: Download build artifacts
uses: ./.github/actions/download-build-artifacts
with:
name: ${{ inputs.build-environment }}
- name: Download TD artifacts
continue-on-error: true
uses: ./.github/actions/download-td-artifacts
- name: Parse ref
id: parse-ref
run: .github/scripts/parse_ref.py
- name: Check for keep-going label and re-enabled test issues
# This uses the filter-test-configs action because it conveniently
# checks for labels and re-enabled test issues. It does not actually do
# any filtering. All filtering is done in the build step.
id: keep-going
uses: ./.github/actions/filter-test-configs
with:
github-token: ${{ secrets.GITHUB_TOKEN }}
test-matrix: ${{ inputs.test-matrix }}
job-name: ${{ steps.get-job-id.outputs.job-name }}
- name: Set Test step time
id: test-timeout
shell: bash
env:
JOB_TIMEOUT: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
run: |
echo "timeout=$((JOB_TIMEOUT-30))" >> "${GITHUB_OUTPUT}"
- name: Preserve github env variables for use in docker
shell: bash
run: |
env | grep '^GITHUB' >> "/tmp/github_env_${GITHUB_RUN_ID}"
env | grep '^CI' >> "/tmp/github_env_${GITHUB_RUN_ID}"
- name: Test
id: test
env:
BUILD_ENVIRONMENT: ${{ inputs.build-environment }}
PR_NUMBER: ${{ github.event.pull_request.number }}
GITHUB_REPOSITORY: ${{ github.repository }}
GITHUB_WORKFLOW: ${{ github.workflow }}
GITHUB_JOB: ${{ github.job }}
GITHUB_RUN_ID: ${{ github.run_id }}
GITHUB_RUN_NUMBER: ${{ github.run_number }}
GITHUB_RUN_ATTEMPT: ${{ github.run_attempt }}
JOB_ID: ${{ steps.get-job-id.outputs.job-id }}
JOB_NAME: ${{ steps.get-job-id.outputs.job-name }}
BRANCH: ${{ steps.parse-ref.outputs.branch }}
SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
BASE_SHA: ${{ github.event.pull_request.base.sha || github.sha }}
TEST_CONFIG: ${{ matrix.config }}
SHARD_NUMBER: ${{ matrix.shard }}
NUM_TEST_SHARDS: ${{ matrix.num_shards }}
REENABLED_ISSUES: ${{ steps.keep-going.outputs.reenabled-issues }}
CONTINUE_THROUGH_ERROR: ${{ steps.keep-going.outputs.keep-going }}
VERBOSE_TEST_LOGS: ${{ steps.keep-going.outputs.ci-verbose-test-logs }}
TEST_SHOWLOCALS: ${{ steps.keep-going.outputs.ci-test-showlocals }}
NO_TEST_TIMEOUT: ${{ steps.keep-going.outputs.ci-no-test-timeout }}
NO_TD: ${{ steps.keep-going.outputs.ci-no-td }}
DOCKER_IMAGE: ${{ inputs.docker-image }}
PYTORCH_TEST_CUDA_MEM_LEAK_CHECK: ${{ matrix.mem_leak_check && '1' || '0' }}
PYTORCH_TEST_RERUN_DISABLED_TESTS: ${{ matrix.rerun_disabled_tests && '1' || '0' }}
TESTS_TO_INCLUDE: ${{ inputs.tests-to-include }}
DASHBOARD_TAG: ${{ inputs.dashboard-tag }}
HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
timeout-minutes: ${{ fromJson(steps.test-timeout.outputs.timeout) }}
run: |
set -x
if [[ $TEST_CONFIG == 'multigpu' ]]; then
TEST_COMMAND=.ci/pytorch/multigpu-test.sh
elif [[ $BUILD_ENVIRONMENT == *onnx* ]]; then
TEST_COMMAND=.ci/caffe2/test.sh
else
TEST_COMMAND=.ci/pytorch/test.sh
fi
# detached container should get cleaned up by teardown_ec2_linux
# TODO: Stop building test binaries as part of the build phase
# Used for GPU_FLAG since that doesn't play nice
# shellcheck disable=SC2086,SC2090
container_name=$(docker run \
${GPU_FLAG:-} \
-e BUILD_ENVIRONMENT \
-e PR_NUMBER \
-e GITHUB_ACTIONS \
-e GITHUB_REPOSITORY \
-e GITHUB_WORKFLOW \
-e GITHUB_JOB \
-e GITHUB_RUN_ID \
-e GITHUB_RUN_NUMBER \
-e GITHUB_RUN_ATTEMPT \
-e JOB_ID \
-e JOB_NAME \
-e BASE_SHA \
-e BRANCH \
-e SHA1 \
-e AWS_DEFAULT_REGION \
-e IN_WHEEL_TEST \
-e SHARD_NUMBER \
-e TEST_CONFIG \
-e NUM_TEST_SHARDS \
-e REENABLED_ISSUES \
-e CONTINUE_THROUGH_ERROR \
-e VERBOSE_TEST_LOGS \
-e TEST_SHOWLOCALS \
-e NO_TEST_TIMEOUT \
-e NO_TD \
-e MAX_JOBS="$(nproc --ignore=2)" \
-e PYTORCH_TEST_CUDA_MEM_LEAK_CHECK \
-e PYTORCH_TEST_RERUN_DISABLED_TESTS \
-e TESTS_TO_INCLUDE \
-e HUGGING_FACE_HUB_TOKEN \
-e DASHBOARD_TAG \
--env-file="${RUNNER_TEMP}/github_env_${GITHUB_RUN_ID}" \
--ulimit stack=10485760:83886080 \
--ulimit core=0 \
--env-file="/tmp/github_env_${GITHUB_RUN_ID}" \
--security-opt seccomp=unconfined \
--cap-add=SYS_PTRACE \
--shm-size="8g" \
--tty \
--detach \
--name="${container_name}" \
--user jenkins \
-v "${GITHUB_WORKSPACE}:/var/lib/jenkins/workspace" \
-w /var/lib/jenkins/workspace \
"${DOCKER_IMAGE}"
)
# save container name for later step
echo "CONTAINER_NAME=${container_name}" >> "$GITHUB_ENV"
# jenkins user does not have write permission to mounted workspace; work-around by copying within container to jenkins home
docker exec -t "${container_name}" sh -c "cd .. && cp -R workspace pytorch && cd pytorch && pip install dist/*.whl && ${TEST_COMMAND}"
- name: Save test results
if: always()
run: |
# copy test results back to the mounted workspace, needed sudo, resulting permissions were correct
docker exec -t "${{ env.CONTAINER_NAME }}" sh -c "cd ../pytorch && sudo cp -R test/test-reports ../workspace/test"
- name: Change permissions (only needed for kubernetes runners for now)
if: ${{ always() && steps.test.conclusion && (contains(matrix.runner, 'gfx942') || contains(matrix.runner, 'mi355')) }}
run: |
docker exec -t "${{ env.CONTAINER_NAME }}" sh -c "sudo chown -R 1001:1001 test"
- name: Print remaining test logs
shell: bash
if: always() && steps.test.conclusion
run: |
cat test/**/*_toprint.log || true
- name: Stop monitoring script
if: ${{ always() && steps.monitor-script.outputs.monitor-script-pid }}
shell: bash
continue-on-error: true
env:
MONITOR_SCRIPT_PID: ${{ steps.monitor-script.outputs.monitor-script-pid }}
run: |
kill "$MONITOR_SCRIPT_PID"
- name: Upload test artifacts
uses: ./.github/actions/upload-test-artifacts
if: always() && steps.test.conclusion && steps.test.conclusion != 'skipped'
with:
use-gha: true
file-suffix: ${{ github.job }}-${{ matrix.config }}-${{ matrix.shard }}-${{ matrix.num_shards }}-${{ matrix.runner }}_${{ steps.get-job-id.outputs.job-id }}
- name: Collect backtraces from coredumps (if any)
if: always()
run: |
# shellcheck disable=SC2156
find . -iname "core.[1-9]*" -exec docker exec "${CONTAINER_NAME}" sh -c "gdb python {} -ex 'bt' -ex 'q'" \;
- name: Store Core dumps on GitHub
uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
if: failure()
with:
name: coredumps-${{ matrix.config }}-${{ matrix.shard }}-${{ matrix.num_shards }}-${{ matrix.runner }}
retention-days: 14
if-no-files-found: ignore
path: ./**/core.[1-9]*
- name: Authenticate with AWS
uses: aws-actions/configure-aws-credentials@ececac1a45f3b08a01d2dd070d28d111c5fe6722 # v4.1.0
with:
role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_upload-benchmark-results
# The max duration enforced by the server side
role-duration-seconds: 18000
aws-region: us-east-1
- name: Upload the benchmark results
uses: pytorch/test-infra/.github/actions/upload-benchmark-results@main
with:
benchmark-results-dir: test/test-reports
dry-run: false
schema-version: v3
github-token: ${{ secrets.GITHUB_TOKEN }}
- name: Upload utilization stats
if: ${{ always() && steps.test.conclusion && steps.test.conclusion != 'skipped' && !inputs.disable-monitor }}
continue-on-error: true
uses: ./.github/actions/upload-utilization-stats
with:
job_id: ${{ steps.get-job-id.outputs.job-id }}
job_name: ${{ steps.get-job-id.outputs.job-name }}
workflow_name: ${{ github.workflow }}
workflow_run_id: ${{github.run_id}}
workflow_attempt: ${{github.run_attempt}}
- name: Teardown ROCm
uses: ./.github/actions/teardown-rocm