pytorch/.github/workflows/_xpu-test.yml
Yang Wang 175ba9fed6 [Utilization Monitor] input to disable utilization monitor (#140857)
# Overview
Currently monitor.py produces error only result, this pr introduct disable-monitor option to all *-test.yml. We also like to explore how the monitor code affect benchmark results.

# next steps
- fix the monitor.py
- enable non-benchmark tests with monitor
- investigate benchmark test behavior with monitor background job

Pull Request resolved: https://github.com/pytorch/pytorch/pull/140857
Approved by: https://github.com/huydhn
2024-11-18 23:26:03 +00:00

291 lines
11 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: xpu-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)
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
env:
GIT_DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
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
timeout-minutes: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
runs-on: ${{ matrix.runner }}
steps:
# [see note: pytorch repo ref]
- name: Checkout PyTorch
uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
- name: Setup XPU
uses: ./.github/actions/setup-xpu
- name: configure aws credentials
id: aws_creds
uses: aws-actions/configure-aws-credentials@v1.7.0
with:
role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_s3_and_ecr_read_only
aws-region: us-east-1
- name: Login to Amazon ECR
id: login-ecr
uses: aws-actions/amazon-ecr-login@v2
- 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: Start monitoring script
id: monitor-script
if: ${{ !inputs.disable-monitor }}
shell: bash
continue-on-error: true
run: |
python3 -m pip install psutil==5.9.1 nvidia-ml-py==11.525.84
python3 -m tools.stats.monitor > usage_log.txt 2>&1 &
echo "monitor-script-pid=${!}" >> "${GITHUB_OUTPUT}"
- name: Download build artifacts
uses: ./.github/actions/download-build-artifacts
with:
name: ${{ inputs.build-environment }}
- name: Parse ref
id: parse-ref
run: .github/scripts/parse_ref.py
- 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: Check for keep-going label and re-enabled test issues
# This uses the filter-test-configs action because it conviniently
# 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: 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 }}
PYTORCH_RETRY_TEST_CASES: 1
PYTORCH_OVERRIDE_FLAKY_SIGNAL: 1
CONTINUE_THROUGH_ERROR: ${{ steps.keep-going.outputs.keep-going }}
TEST_SHOWLOCALS: ${{ steps.keep-going.outputs.ci-test-showlocals }}
VERBOSE_TEST_LOGS: ${{ steps.keep-going.outputs.ci-verbose-test-logs }}
NO_TEST_TIMEOUT: ${{ steps.keep-going.outputs.ci-no-test-timeout }}
NO_TD: ${{ steps.keep-going.outputs.ci-no-td }}
TEST_CONFIG: ${{ matrix.config }}
SHARD_NUMBER: ${{ matrix.shard }}
NUM_TEST_SHARDS: ${{ matrix.num_shards }}
REENABLED_ISSUES: ${{ steps.keep-going.outputs.reenabled-issues }}
SCCACHE_BUCKET: ossci-compiler-cache-circleci-v2
SCCACHE_REGION: us-east-1
SCCACHE_S3_KEY_PREFIX: ${{ github.workflow }}
DOCKER_IMAGE: ${{ inputs.docker-image }}
XLA_CLANG_CACHE_S3_BUCKET_NAME: ossci-compiler-clang-cache-circleci-xla
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 }}
timeout-minutes: ${{ fromJson(steps.test-timeout.outputs.timeout) }}
run: |
# Fetch aws credential from IMDs
eval "$(python3 .github/scripts/get_aws_session_tokens.py)"
set -x
TEST_COMMAND=.ci/pytorch/test.sh
# detached container should get cleaned up by teardown_ec2_linux
# Used for GPU_FLAG since that doesn't play nice
# shellcheck disable=SC2086,SC2090
container_name=$(docker run \
${GPU_FLAG:-} \
-e BUILD_ENVIRONMENT \
-e PR_NUMBER \
-e GITHUB_ACTIONS \
-e GITHUB_REPOSITORY \
-e GITHUB_WORKFLOW \
-e GITHUB_JOB \
-e GITHUB_RUN_ID \
-e GITHUB_RUN_NUMBER \
-e GITHUB_RUN_ATTEMPT \
-e JOB_ID \
-e BRANCH \
-e SHA1 \
-e AWS_DEFAULT_REGION \
-e AWS_ACCESS_KEY_ID \
-e AWS_SECRET_ACCESS_KEY \
-e AWS_SESSION_TOKEN \
-e IN_WHEEL_TEST \
-e SHARD_NUMBER \
-e TEST_CONFIG \
-e NUM_TEST_SHARDS \
-e REENABLED_ISSUES \
-e PYTORCH_RETRY_TEST_CASES \
-e PYTORCH_OVERRIDE_FLAKY_SIGNAL \
-e CONTINUE_THROUGH_ERROR \
-e VERBOSE_TEST_LOGS \
-e TEST_SHOWLOCALS \
-e NO_TEST_TIMEOUT \
-e NO_TD \
-e MAX_JOBS="$(nproc --ignore=2)" \
-e SCCACHE_BUCKET \
-e SCCACHE_REGION \
-e SCCACHE_S3_KEY_PREFIX \
-e XLA_CLANG_CACHE_S3_BUCKET_NAME \
-e PYTORCH_TEST_CUDA_MEM_LEAK_CHECK \
-e PYTORCH_TEST_RERUN_DISABLED_TESTS \
-e TESTS_TO_INCLUDE \
-e ZE_AFFINITY_MASK \
--env-file="/tmp/github_env_${GITHUB_RUN_ID}" \
--ulimit stack=10485760:83886080 \
--ulimit core=0 \
--security-opt seccomp=unconfined \
--cap-add=SYS_PTRACE \
--shm-size="8g" \
--tty \
--detach \
--name="${container_name}" \
--user jenkins \
--privileged \
-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: 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 "${CONTAINER_NAME}" sh -c "gdb python {} -ex 'bt' -ex 'q'" \;
- name: Stop container before exit
if: always()
run: |
# Workaround for multiple runners on same IDC node
docker stop "${{ env.CONTAINER_NAME }}"
- name: Store Core dumps on GitHub
uses: actions/upload-artifact@v4
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: Teardown XPU
uses: ./.github/actions/teardown-xpu