exo/.github/workflows/benchmarks.yml

name: Build and Test

on:
  push:
    branches: [ '*' ]
    tags: [ '*' ]
  pull_request:
    branches: [ '*' ]

jobs:
  single-m4-pro:
    strategy:
      matrix:
        model: ['llama-3.2-1b', 'llama-3.2-3b', 'llama-3.1-8b']
    uses: ./.github/workflows/bench_job.yml
    with:
      config: '{"M4PRO_GPU16_24GB": 1}'
      model: ${{ matrix.model }}
      calling_job_name: 'single-m4-pro'
      network_interface: 'Ethernet'
    secrets: inherit

  two-m4-pro-cluster:
    strategy:
      matrix:
        model: ['llama-3.2-1b', 'llama-3.2-3b', 'llama-3.1-8b']
    uses: ./.github/workflows/bench_job.yml
    with:
      config: '{"M4PRO_GPU16_24GB": 2}'
      model: ${{ matrix.model }}
      calling_job_name: 'two-m4-pro-cluster'
      network_interface: 'Ethernet'
    secrets: inherit

  # two-m4-pro-cluster-thunderbolt:
  #   strategy:
  #     matrix:
  #       model: ['llama-3.2-1b', 'llama-3.2-3b', 'llama-3.1-8b']
  #   uses: ./.github/workflows/bench_job.yml
  #   with:
  #     config: '{"M4PRO_GPU16_24GB": 2}'
  #     model: ${{ matrix.model }}
  #     calling_job_name: 'two-m4-pro-cluster-thunderbolt'
  #     network_interface: 'Thunderbolt'
  #   secrets: inherit

  three-m4-pro-cluster:
    strategy:
      matrix:
        model: ['llama-3.2-1b', 'llama-3.2-3b', 'llama-3.1-8b', 'llama-3.3-70b']
      fail-fast: false
    uses: ./.github/workflows/bench_job.yml
    with:
      config: '{"M4PRO_GPU16_24GB": 3}'
      model: ${{ matrix.model }}
      calling_job_name: 'three-m4-pro-cluster'
      network_interface: 'Ethernet'
    secrets: inherit

  # test-m3-single-node:
  #   strategy:
  #     matrix:
  #       model: ['llama-3.2-1b']
  #     fail-fast: false
  #   uses: ./.github/workflows/bench_job.yml
  #   with:
  #     config: '{"M3MAX_GPU40_128GB": 1}'
  #     model: ${{ matrix.model }}
  #     calling_job_name: 'test-m3-cluster'
  #     network_interface: 'Ethernet'
  #   secrets: inherit