dvc-studio.yml

name: DVC Studio Experiment

on:

  push:
    tags-ignore:
      - '**'

  workflow_dispatch:
    inputs:
      exp-run-args:
        description: 'Args to be passed to dvc exp run call'
        required: false
        type: string
        default: ''
      parent-sha:
        description: 'SHA of the commit to start the experiment from'
        required: false
        type: string
        default: ''
      cloud:
        description: 'Cloud compute provider to host the runner'
        required: false
        default: 'aws'
        type: choice
        options:
          - aws
          - azure
          - gcp
      type:
        description: 'https://registry.terraform.io/providers/iterative/iterative/latest/docs/resources/task#machine-type'
        required: false
        default: 'g5.2xlarge'
      region:
        description: 'https://registry.terraform.io/providers/iterative/iterative/latest/docs/resources/task#cloud-region'
        required: false
        default: 'us-east'
      spot:
        description: 'Request a spot instance'
        required: false
        default: false
        type: boolean
      storage:
        description: 'Disk size in GB'
        required: false
        default: 40
        type: number
      timeout:
        description: 'Timeout in seconds'
        required: false
        default: 3600
        type: number
  
permissions:
  contents: write
  id-token: write
  pull-requests: write

jobs:

  deploy-runner:
    if: ${{ (github.actor == 'iterative-studio[bot]') || (github.event_name == 'workflow_dispatch') }} 
    environment: cloud
    runs-on: ubuntu-latest

    steps:
      - uses: actions/checkout@v3
        with:
          ref: ${{ inputs.parent-sha || '' }}
      - uses: iterative/setup-cml@v2
      - uses: aws-actions/configure-aws-credentials@v4
        with:
          aws-region: us-east-2
          role-to-assume: ${{ vars.AWS_SANDBOX_ROLE }}
          role-duration-seconds: 43200
      - name: Create Runner
        env:
          REPO_TOKEN: ${{ secrets.PERSONAL_ACCESS_TOKEN }}
        run: |
          cml runner launch --single \
            --labels=cml \
            --cloud=${{ inputs.cloud || 'aws' }} \
            --cloud-region=${{ inputs.region || 'us-east' }} \
            --cloud-hdd-size=${{ inputs.storage || '40' }} \
            --cloud-type=${{ inputs.type || 'g5.2xlarge' }} \
            --idle-timeout=${{ inputs.timeout || '3600' }} \
             ${{ (inputs.spot == 'true' && '--cloud-spot') || '' }}

  runner-job:
    needs: deploy-runner
    runs-on: [ self-hosted, cml ]
    environment: cloud
    container:
      image: iterativeai/cml:latest-gpu
      options: --gpus all --ipc host

    steps:
      - uses: actions/checkout@v3
        with:
          ref: ${{ inputs.parent-sha || '' }}
      - uses: aws-actions/configure-aws-credentials@v4
        with:
          aws-region: us-east-2
          role-to-assume: ${{ vars.AWS_SANDBOX_ROLE }}
          role-duration-seconds: 43200

      - run: pip install -r requirements.txt

      - name: Train
        env:
          REPO_TOKEN: ${{ secrets.PERSONAL_ACCESS_TOKEN }}
          DVC_STUDIO_TOKEN: ${{ secrets.DVC_STUDIO_TOKEN }}
          DVCLIVE_LOGLEVEL: DEBUG
        run: |
          cml ci --fetch-depth 0
          dvc exp run --pull --allow-missing ${{ github.event.inputs.exp-run-args }}
          dvc remote add --local push_remote s3://dvc-public/remote/get-started-pools 

      - name: Workflow Dispatch Sharing
        if: github.event_name == 'workflow_dispatch'
        env:
          DVC_STUDIO_TOKEN: ${{ secrets.DVC_STUDIO_TOKEN }}
        run: |
          dvc exp push origin -r push_remote

      - name: Commit-based Sharing
        if: github.actor == 'iterative-studio[bot]'
        env:
          REPO_TOKEN: ${{ secrets.PERSONAL_ACCESS_TOKEN }}
        run: |
          dvc push -r push_remote
          cml pr --squash --skip-ci .
          echo "## Metrics" > report.md
          dvc metrics diff main --md >> report.md
          echo "## Params" >> report.md
          dvc params diff main --md >> report.md
          cml comment create --pr report.md