Skip to content

inductor-A100-perf-nightly #12

inductor-A100-perf-nightly

inductor-A100-perf-nightly #12

name: inductor-A100-perf-nightly
on:
schedule:
- cron: 0 7 * * *
workflow_dispatch:
inputs:
training:
description: Run training?
required: false
type: boolean
default: true
inference:
description: Run inference?
required: false
type: boolean
default: false
default:
description: Run inductor_default?
required: false
type: boolean
default: true
dynamic:
description: Run inductor_dynamic_shapes?
required: false
type: boolean
default: true
cudagraphs:
description: Run inductor_cudagraphs?
required: false
type: boolean
default: true
cppwrapper:
description: Run inductor_cpp_wrapper for inference?
required: false
type: boolean
default: false
aotinductor:
description: Run aot_inductor for inference?
required: false
type: boolean
default: false
maxautotune:
description: Run inductor_max_autotune?
required: false
type: boolean
default: false
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
jobs:
linux-bionic-cuda11_8-py3_10-gcc7-inductor-build:
name: cuda11.8-py3.10-gcc7-sm80
uses: ./.github/workflows/_linux-build.yml
with:
build-environment: linux-bionic-cuda11.8-py3.10-gcc7-sm80
docker-image-name: pytorch-linux-bionic-cuda11.8-cudnn8-py3-gcc7-inductor-benchmarks
cuda-arch-list: '8.0'
test-matrix: |
{ include: [
{ config: "inductor_huggingface_perf", shard: 1, num_shards: 3, runner: "linux.gcp.a100.large" },
{ config: "inductor_huggingface_perf", shard: 2, num_shards: 3, runner: "linux.gcp.a100.large" },
{ config: "inductor_huggingface_perf", shard: 3, num_shards: 3, runner: "linux.gcp.a100.large" },
{ config: "inductor_timm_perf", shard: 1, num_shards: 5, runner: "linux.gcp.a100.large" },
{ config: "inductor_timm_perf", shard: 2, num_shards: 5, runner: "linux.gcp.a100.large" },
{ config: "inductor_timm_perf", shard: 3, num_shards: 5, runner: "linux.gcp.a100.large" },
{ config: "inductor_timm_perf", shard: 4, num_shards: 5, runner: "linux.gcp.a100.large" },
{ config: "inductor_timm_perf", shard: 5, num_shards: 5, runner: "linux.gcp.a100.large" },
{ config: "inductor_torchbench_perf", shard: 1, num_shards: 4, runner: "linux.gcp.a100.large" },
{ config: "inductor_torchbench_perf", shard: 2, num_shards: 4, runner: "linux.gcp.a100.large" },
{ config: "inductor_torchbench_perf", shard: 3, num_shards: 4, runner: "linux.gcp.a100.large" },
{ config: "inductor_torchbench_perf", shard: 4, num_shards: 4, runner: "linux.gcp.a100.large" },
]}
linux-bionic-cuda11_8-py3_10-gcc7-inductor-test-nightly:
name: cuda11.8-py3.10-gcc7-sm80
uses: ./.github/workflows/_linux-test.yml
needs: linux-bionic-cuda11_8-py3_10-gcc7-inductor-build
if: github.event.schedule == '0 7 * * *'
with:
build-environment: linux-bionic-cuda11.8-py3.10-gcc7-sm80
dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-cppwrapper-true
docker-image: ${{ needs.linux-bionic-cuda11_8-py3_10-gcc7-inductor-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-bionic-cuda11_8-py3_10-gcc7-inductor-build.outputs.test-matrix }}
use-gha: anything-non-empty-to-use-gha
timeout-minutes: 720
linux-bionic-cuda11_8-py3_10-gcc7-inductor-test:
name: cuda11.8-py3.10-gcc7-sm80
uses: ./.github/workflows/_linux-test.yml
needs: linux-bionic-cuda11_8-py3_10-gcc7-inductor-build
if: github.event_name == 'workflow_dispatch'
with:
build-environment: linux-bionic-cuda11.8-py3.10-gcc7-sm80
dashboard-tag: training-${{ inputs.training }}-inference-${{ inputs.inference }}-default-${{ inputs.default }}-dynamic-${{ inputs.dynamic }}-cudagraphs-${{ inputs.cudagraphs }}-cppwrapper-${{ inputs.cppwrapper }}-aotinductor-${{ inputs.aotinductor }}-maxautotune-${{ inputs.maxautotune }}
docker-image: ${{ needs.linux-bionic-cuda11_8-py3_10-gcc7-inductor-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-bionic-cuda11_8-py3_10-gcc7-inductor-build.outputs.test-matrix }}
use-gha: anything-non-empty-to-use-gha
timeout-minutes: 720