inductor-perf-nightly-aarch64 #336
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: inductor-perf-nightly-aarch64 | |
| on: | |
| schedule: | |
| # - cron: 0 7 * * 1-6 | |
| # - cron: 0 7 * * 0 | |
| # Does not perform max_autotune on CPU, so skip the weekly run setup | |
| # Run 6 times everyday to see if perf instablity can be reproduced | |
| # Will change this back | |
| - cron: 0 */4 * * * | |
| # NB: GitHub has an upper limit of 10 inputs here | |
| workflow_dispatch: | |
| inputs: | |
| training: | |
| # CPU for training is not typical, but leave the option open here | |
| description: Run training (off by default)? | |
| required: false | |
| type: boolean | |
| default: false | |
| inference: | |
| description: Run inference (on by default)? | |
| required: false | |
| type: boolean | |
| default: true | |
| default: | |
| description: Run inductor_default? | |
| required: false | |
| type: boolean | |
| default: true | |
| dynamic: | |
| description: Run inductor_dynamic_shapes? | |
| required: false | |
| type: boolean | |
| default: false | |
| aotinductor: | |
| description: Run aot_inductor for inference? | |
| required: false | |
| type: boolean | |
| default: false | |
| benchmark_configs: | |
| description: The list of configs used the benchmark | |
| required: false | |
| type: string | |
| default: inductor_huggingface_perf_cpu_aarch64,inductor_timm_perf_cpu_aarch64,inductor_torchbench_perf_cpu_aarch64 | |
| concurrency: | |
| group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ github.ref_type == 'branch' && github.sha }}-${{ github.event_name == 'workflow_dispatch' }}-${{ github.event_name == 'schedule' }} | |
| cancel-in-progress: true | |
| permissions: read-all | |
| jobs: | |
| linux-jammy-aarch64-py3_10-inductor-build: | |
| name: linux-jammy-aarch64-py3.10-inductor | |
| uses: ./.github/workflows/_linux-build.yml | |
| with: | |
| runner: linux.arm64.m7g.4xlarge | |
| build-environment: linux-jammy-aarch64-py3.10 | |
| docker-image-name: pytorch-linux-jammy-aarch64-py3.10-gcc11-inductor-benchmarks | |
| test-matrix: | | |
| { include: [ | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 1, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 2, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 3, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 4, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 5, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 6, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 7, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 8, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_huggingface_perf_cpu_aarch64", shard: 9, num_shards: 9, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 1, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 2, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 3, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 4, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 5, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 6, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 7, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 8, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 9, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 10, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 11, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 12, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 13, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 14, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_timm_perf_cpu_aarch64", shard: 15, num_shards: 15, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 1, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 2, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 3, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 4, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 5, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 6, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 7, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 8, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 9, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 10, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 11, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| { config: "inductor_torchbench_perf_cpu_aarch64", shard: 12, num_shards: 12, runner: "linux.arm64.m7g.metal" }, | |
| ]} | |
| selected-test-configs: ${{ inputs.benchmark_configs }} | |
| secrets: | |
| HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }} | |
| linux-jammy-aarch64-py3_10-inductor-test-nightly: | |
| name: linux-jammy-aarch64-py3.10-inductor | |
| uses: ./.github/workflows/_linux-test.yml | |
| needs: linux-jammy-aarch64-py3_10-inductor-build | |
| if: github.event.schedule == '0 */4 * * *' | |
| with: | |
| build-environment: linux-jammy-aarch64-py3.10 | |
| # Turn off dynamic-shapes and aotinductor tests for now, to have faster iteration for debugging perf instability. | |
| # Will change this back | |
| dashboard-tag: training-false-inference-true-default-true-dynamic-false-aotinductor-false | |
| docker-image: ${{ needs.linux-jammy-aarch64-py3_10-inductor-build.outputs.docker-image }} | |
| test-matrix: ${{ needs.linux-jammy-aarch64-py3_10-inductor-build.outputs.test-matrix }} | |
| use-gha: anything-non-empty-to-use-gha | |
| timeout-minutes: 720 | |
| secrets: | |
| HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }} | |
| linux-jammy-aarch64-py3_10-inductor-test: | |
| name: linux-jammy-aarch64-py3.10-inductor | |
| uses: ./.github/workflows/_linux-test.yml | |
| needs: linux-jammy-aarch64-py3_10-inductor-build | |
| if: github.event_name == 'workflow_dispatch' | |
| with: | |
| build-environment: linux-jammy-aarch64-py3.10 | |
| dashboard-tag: training-${{ inputs.training }}-inference-${{ inputs.inference }}-default-${{ inputs.default }}-dynamic-${{ inputs.dynamic }}-aotinductor-${{ inputs.aotinductor }} | |
| docker-image: ${{ needs.linux-jammy-aarch64-py3_10-inductor-build.outputs.docker-image }} | |
| test-matrix: ${{ needs.linux-jammy-aarch64-py3_10-inductor-build.outputs.test-matrix }} | |
| use-gha: anything-non-empty-to-use-gha | |
| timeout-minutes: 720 | |
| secrets: | |
| HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }} |