mirror of
https://github.com/pytorch/pytorch.git
synced 2025-10-20 21:14:14 +08:00
465fe4d9f7
Per the discussion with @nWEIdia, this resumes the work on https://github.com/pytorch/pytorch/pull/157870 to enable PT2 benchmark on B200 ### Testing https://github.com/pytorch/pytorch/actions/runs/16615101382 Pull Request resolved: https://github.com/pytorch/pytorch/pull/158011 Approved by: https://github.com/nWEIdia, https://github.com/atalman
155 lines
6.2 KiB
YAML
155 lines
6.2 KiB
YAML
name: inductor-perf-b200
|
|
|
|
on:
|
|
schedule:
|
|
- cron: 0 7 * * 1-6
|
|
- cron: 0 7 * * 0
|
|
# NB: GitHub has an upper limit of 10 inputs here, so before we can sort it
|
|
# out, let try to run torchao cudagraphs_low_precision as part of cudagraphs
|
|
workflow_dispatch:
|
|
inputs:
|
|
training:
|
|
description: Run training (on by default)?
|
|
required: false
|
|
type: boolean
|
|
default: true
|
|
inference:
|
|
description: Run inference (on by default)?
|
|
required: false
|
|
type: boolean
|
|
default: true
|
|
default:
|
|
description: Run inductor_default?
|
|
required: false
|
|
type: boolean
|
|
default: false
|
|
dynamic:
|
|
description: Run inductor_dynamic_shapes?
|
|
required: false
|
|
type: boolean
|
|
default: false
|
|
cppwrapper:
|
|
description: Run inductor_cpp_wrapper?
|
|
required: false
|
|
type: boolean
|
|
default: false
|
|
cudagraphs:
|
|
description: Run inductor_cudagraphs?
|
|
required: false
|
|
type: boolean
|
|
default: true
|
|
freezing_cudagraphs:
|
|
description: Run inductor_cudagraphs with freezing 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
|
|
benchmark_configs:
|
|
description: The list of configs used the benchmark
|
|
required: false
|
|
type: string
|
|
default: inductor_huggingface_perf_cuda_b200,inductor_timm_perf_cuda_b200,inductor_torchbench_perf_cuda_b200
|
|
|
|
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:
|
|
id-token: write
|
|
contents: read
|
|
|
|
jobs:
|
|
get-label-type:
|
|
name: get-label-type
|
|
uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
|
|
if: ${{ (github.event_name != 'schedule' || github.repository == 'pytorch/pytorch') && github.repository_owner == 'pytorch' }}
|
|
with:
|
|
triggering_actor: ${{ github.triggering_actor }}
|
|
issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
|
|
curr_branch: ${{ github.head_ref || github.ref_name }}
|
|
curr_ref_type: ${{ github.ref_type }}
|
|
opt_out_experiments: lf
|
|
|
|
build:
|
|
name: cuda12.8-py3.10-gcc9-sm100
|
|
uses: ./.github/workflows/_linux-build.yml
|
|
needs: get-label-type
|
|
with:
|
|
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
|
|
# Use a bigger runner here because CUDA_ARCH 9.0 is only built for H100
|
|
# or newer GPUs, so it doesn't benefit much from existing compiler cache
|
|
# from trunk. Also use a memory-intensive runner here because memory is
|
|
# usually the bottleneck
|
|
runner: linux.12xlarge.memory
|
|
build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm100
|
|
docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
|
|
cuda-arch-list: '10.0'
|
|
test-matrix: |
|
|
{ include: [
|
|
{ config: "inductor_huggingface_perf_cuda_b200", shard: 1, num_shards: 1, runner: "linux.dgx.b200" },
|
|
{ config: "inductor_timm_perf_cuda_b200", shard: 1, num_shards: 1, runner: "linux.dgx.b200" },
|
|
{ config: "inductor_torchbench_perf_cuda_b200", shard: 1, num_shards: 1, runner: "linux.dgx.b200" },
|
|
]}
|
|
selected-test-configs: ${{ inputs.benchmark_configs }}
|
|
build-additional-packages: "vision audio fbgemm torchao"
|
|
secrets: inherit
|
|
|
|
test-periodically:
|
|
name: cuda12.8-py3.10-gcc9-sm100
|
|
uses: ./.github/workflows/_linux-test.yml
|
|
needs: build
|
|
if: github.event.schedule == '0 7 * * 1-6'
|
|
with:
|
|
build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm100
|
|
dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-cppwrapper-true-aotinductor-true-freezing_cudagraphs-true-cudagraphs_low_precision-true
|
|
docker-image: ${{ needs.build.outputs.docker-image }}
|
|
test-matrix: ${{ needs.build.outputs.test-matrix }}
|
|
aws-role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_s3_and_ecr_read_only
|
|
timeout-minutes: 720
|
|
disable-monitor: false
|
|
monitor-log-interval: 15
|
|
monitor-data-collect-interval: 4
|
|
secrets: inherit
|
|
|
|
test-weekly:
|
|
name: cuda12.8-py3.10-gcc9-sm100
|
|
uses: ./.github/workflows/_linux-test.yml
|
|
needs: build
|
|
if: github.event.schedule == '0 7 * * 0'
|
|
with:
|
|
build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm100
|
|
dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-cppwrapper-true-aotinductor-true-freezing_cudagraphs-true-maxautotune-true-freeze_autotune_cudagraphs-true-cudagraphs_low_precision-true
|
|
docker-image: ${{ needs.build.outputs.docker-image }}
|
|
test-matrix: ${{ needs.build.outputs.test-matrix }}
|
|
timeout-minutes: 1440
|
|
aws-role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_s3_and_ecr_read_only
|
|
disable-monitor: false
|
|
monitor-log-interval: 15
|
|
monitor-data-collect-interval: 4
|
|
secrets: inherit
|
|
|
|
test:
|
|
name: cuda12.8-py3.10-gcc9-sm100
|
|
uses: ./.github/workflows/_linux-test.yml
|
|
needs: build
|
|
with:
|
|
build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm100
|
|
dashboard-tag: training-${{ inputs.training }}-inference-${{ inputs.inference }}-default-${{ inputs.default }}-dynamic-${{ inputs.dynamic }}-cudagraphs-${{ inputs.cudagraphs }}-cppwrapper-${{ inputs.cppwrapper }}-aotinductor-${{ inputs.aotinductor }}-maxautotune-${{ inputs.maxautotune }}-freezing_cudagraphs-${{ inputs.freezing_cudagraphs }}-cudagraphs_low_precision-${{ inputs.cudagraphs }}
|
|
docker-image: ${{ needs.build.outputs.docker-image }}
|
|
test-matrix: ${{ needs.build.outputs.test-matrix }}
|
|
aws-role-to-assume: arn:aws:iam::308535385114:role/gha_workflow_s3_and_ecr_read_only
|
|
timeout-minutes: 720
|
|
disable-monitor: false
|
|
monitor-log-interval: 15
|
|
monitor-data-collect-interval: 4
|
|
secrets: inherit
|