nm-build.yml

name: nm build
on:
  # makes workflow reusable
  workflow_call:
    inputs:
      wf_category:
        description: "categories: REMOTE, NIGHTLY, RELEASE"
        type: string
        default: "REMOTE"
      build_label:
        description: "requested runner label (specifies instance)"
        type: string
        required: true
      timeout:
        description: "time limit for run in minutes "
        type: string
        required: true
      gitref:
        description: "git commit hash or branch name"
        type: string
        required: true
      Gi_per_thread:
        description: 'requested GiB to reserve per thread'
        type: string
        required: true
      nvcc_threads:
        description: 'requested number of threads for nvcc'
        type: string
        required: true
      python:
        description: "python version, e.g. 3.10.12"
        type: string
        required: true
    outputs:
      whl:
        description: 'basename for generated whl'
        value: ${{ jobs.BUILD.outputs.whl }}

  # makes workflow manually callable
  workflow_dispatch:
    inputs:
      wf_category:
        description: "categories: REMOTE, NIGHTLY, RELEASE"
        type: string
        default: "REMOTE"
      build_label:
        description: "requested runner label (specifies instance)"
        type: string
        required: true
      timeout:
        description: "time limit for run in minutes "
        type: string
        required: true
      gitref:
        description: "git commit hash or branch name"
        type: string
        required: true
      Gi_per_thread:
        description: 'requested GiB to reserve per thread'
        type: string
        required: true
      nvcc_threads:
        description: 'requested number of threads for nvcc'
        type: string
        required: true
      python:
        description: "python version, e.g. 3.10.12"
        type: string
        required: true

env:
    VENV_BASE: "BUILD"

jobs:

    BUILD:

        runs-on: ${{ inputs.build_label }}
        timeout-minutes: ${{ fromJson(inputs.timeout) }}

        permissions:
            contents: 'read'
            id-token: 'write'

        outputs:
            run_id: ${{ github.run_id }}
            whl: ${{ steps.build.outputs.whl }}
            tarfile: ${{ steps.build.outputs.tarfile }}

        steps:

            - name: checkout
              id: checkout
              uses: actions/checkout@v4
              with:
                fetch-depth: 0
                ref: ${{ inputs.gitref }}
                submodules: recursive

            - name: setenv
              id: setenv
              uses: ./.github/actions/nm-set-env/
              with:
                wf_category: ${{ inputs.wf_category }}
                hf_token: ${{ secrets.NM_HF_TOKEN }}
                Gi_per_thread: ${{ inputs.Gi_per_thread }}
                nvcc_threads: ${{ inputs.nvcc_threads }}

            - name: set python
              id: set_python
              uses: neuralmagic/nm-actions/actions/set-python@v1.0.0
              with:
                python: ${{ inputs.python }}
                venv: ${{ env.VENV_BASE }}

            - name: create testmo run
              id: create_testmo_run
              uses: ./.github/actions/nm-testmo-run-create/
              if: success() || failure()
              with:
                testmo_url: https://neuralmagic.testmo.net
                testmo_token: ${{ secrets.TESTMO_TEST_TOKEN }}
                source: 'build-test'

            - name: build
              id: build
              uses: ./.github/actions/nm-build-vllm/
              with:
                python: ${{ inputs.python }}
                venv: ${{ env.VENV_BASE }}

            # GCP
            - name: 'Authenticate to Google Cloud'
              id: auth
              uses: google-github-actions/auth@v2.1.3
              with:
                  project_id: ${{ secrets.GCP_PROJECT }}
                  workload_identity_provider: ${{ secrets.GCP_WORKLOAD_IDENTITY_PROVIDER }}
                  service_account: ${{ secrets.NM_PYPI_SA }}

            - name: 'Set up Cloud SDK'
              uses: 'google-github-actions/setup-gcloud@v2'
              with:
                  version: '>= 473.0.0'

            - name: copy whl and source distribution
              run: |
                # echo "whl: ${{ steps.build.outputs.whl }}"
                # echo "tarfile: ${{ steps.build.outputs.tarfile }}"
                gcloud storage cp dist/${{ steps.build.outputs.whl }} gs://neuralmagic-public-pypi/assets/${{ github.run_id }}/${{ steps.build.outputs.whl }}
                gcloud storage cp dist/${{ steps.build.outputs.tarfile }} gs://neuralmagic-public-pypi/assets/${{ github.run_id }}/${{ steps.build.outputs.tarfile }}

            - name: upload whl
              uses: actions/upload-artifact@v4
              if: success() || failure()
              with:
                name: ${{ steps.build.outputs.whl }}
                path: dist/${{ steps.build.outputs.whl }}
                retention-days: 5

            - name: upload tar.gz
              uses: actions/upload-artifact@v4
              if: success() || failure()
              with:
                name: ${{ inputs.python }}-${{ steps.build.outputs.tarfile }}
                path: dist/${{ steps.build.outputs.tarfile }}
                retention-days: 5

            - name: summary
              uses: ./.github/actions/nm-summary-build/
              if: success() || failure()
              with:
                label: ${{ inputs.build_label }}
                gitref: ${{ inputs.gitref }}
                testmo_run_url: https://neuralmagic.testmo.net/automation/runs/view/${{ steps.create_testmo_run.outputs.id }}
                python: ${{ steps.set_python.outputs.version }}
                whl_status: ${{ steps.build.outputs.whl_status }}

            - name: run status
              id: run_status
              if: success() || failure()
              env:
                BUILD_STATUS: ${{ steps.build.outputs.build_status }}
                WHL_STATUS: ${{ steps.build.outputs.whl_status }}
              run: |
                echo "build status: ${WHL_STATUS}"
                if [ -z "${WHL_STATUS}" ] || [ "${WHL_STATUS}" -ne "0" ]; then exit 1; fi

            - name: complete testmo run
              uses: ./.github/actions/nm-testmo-run-complete/
              if: success() || failure()
              with:
                testmo_url: https://neuralmagic.testmo.net
                testmo_token: ${{ secrets.TESTMO_TEST_TOKEN }}
                testmo_run_id: ${{ steps.create_testmo_run.outputs.id }}