funasr-quick-start.yml

# FunASR quick start guard - project thin trigger.
#
# Calls the common engine .github/workflows/quick-start-template.yml;
# this file only declares what varies per project: schedule,
# concurrency, runner, container image, upstream repo, monitored doc
# and the test entry command. The guard loop, cache I/O, state relay
# and publishing all live in the engine. See the engine's
# `workflow_call.inputs` block for the full input contract.
#
# Model downloads land in /root/.cache/modelscope: no explicit bind
# mount is needed because NFS already mounts /root/.cache (same setup
# as the modelscope / peft / comfyui triggers).

name: funasr-quick-start

concurrency:
  # format() is load-bearing: a '||' between 'manual-' and
  # github.run_id would short-circuit on the truthy literal and
  # every dispatch would share one 'manual-' group.
  group: ${{ github.event_name == 'schedule' && 'funasr-quick-start-schedule' || format('manual-{0}', github.run_id) }}
  # cancel-in-progress: false because (1) a schedule run cancelled
  # mid-way loses its outcome writeback - the outcome is what makes
  # the retry mechanism work, and the 'if: always()' guard isn't
  # enough when the container is being torn down; (2) dispatch / PR
  # runs already live in unique groups so there's nothing to cancel.
  cancel-in-progress: false

on:
  schedule:
    # Daily: '30 4 * * *' fires at 04:30 UTC (12:30 北京时间). FunASR
    # releases on its own schedule, so a once-a-day audit is enough for
    # doc fixes to propagate within a day.
    - cron: '30 4 * * *'
  workflow_dispatch:
  # PR trigger: docs/tests changes get a guard run. paths filter avoids burning
  # the self-hosted NPU runner on unrelated PRs. `pull_request` (not
  # `pull_request_target`): contents: read is enough, no write-token risk.
  pull_request:
    branches: [main]
    paths:
      - 'sources/funasr/**'
      - 'tests/funasr/**'

permissions:
  contents: read

jobs:
  funasr-quick-start:
    uses: ./.github/workflows/quick-start-template.yml
    with:
      # Namespaces cache keys (monitor-state-funasr-*), artifacts
      # (funasr-quick-start-<run_id>) and the test working dir
      # (tests/funasr).
      project: funasr
      # Self-hosted NPU runner for the test job only; the engine pins
      # the cache I/O jobs (restore-cache / publish-and-persist) to
      # GitHub-hosted ubuntu-latest.
      test_runner: '["linux-aarch64-a2-1"]'
      image: swr.cn-south-1.myhuaweicloud.com/ascendhub/cann:9.1.0-910b-ubuntu22.04-py3.12
      # 4 h budget: cold run reuses the image torch stack + source
      # installs funasr at the latest release tag (git clone + uv pip
      # install -e .) + pulls ~3.3 GB of FunASR models from ModelScope
      # (paraformer-zh / SenseVoiceSmall / fsmn-vad / ct-punc / cam++ /
      # paraformer-zh-streaming / emotion2vec_plus_large, warm after the
      # first run via the NFS-mounted /root/.cache) + 5 NPU smokes. The
      # cold download is throttled to <1 MB/s on some ModelScope CDN
      # nodes, so the total budget must cover several 1 GB checkpoints.
      timeout_minutes: 240
      upstream_repo: modelscope/FunASR
      # Doc URL points to the upstream Ascend/docs repo. {0} is filled by the
      # engine: PR head SHA on PR runs, 'main' otherwise.
      doc_url: 'https://raw.githubusercontent.com/Ascend/docs/{0}/sources/funasr/quick_start.md'
      doc_path: https://github.com/Ascend/docs/blob/main/sources/funasr/quick_start.md
      # cwd is the repo root inside the checkout; env contract
      # (MONITORED_DOC_URL / UPSTREAM_REF / NPU_READY) is injected by the
      # engine. All project env prep and installs live in the test
      # subclass's prepare_environment hook.
      test_command: python -m unittest tests.funasr.test_quick_start_ascend -v 2>&1