forked from pytorch/pytorch
-
Notifications
You must be signed in to change notification settings - Fork 0
136 lines (129 loc) · 6.4 KB
/
inductor-perf-test-nightly-a10g.yml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
name: inductor-perf-nightly-A10g
on:
schedule:
# - cron: 0 7 * * 1-6
# - cron: 0 7 * * 0
# Do not perform weekly max-autotune run for now.
- cron: 0 7 * * *
# 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 (off by default)?
required: false
type: boolean
default: false
default:
description: Run inductor_default?
required: false
type: boolean
default: false
dynamic:
description: Run inductor_dynamic_shapes?
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
freeze_autotune_cudagraphs:
description: Run inductor_cudagraphs with freezing and max autotune 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_a10g,inductor_timm_perf_cuda_a10g,inductor_torchbench_perf_cuda_a10g
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:
get-label-type:
name: get-label-type
uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
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: ${{ github.ref_type }}
linux-focal-cuda12_1-py3_10-gcc9-inductor-build:
name: cuda12.1-py3.10-gcc9-sm80
uses: ./.github/workflows/_linux-build.yml
needs: get-label-type
with:
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
build-environment: linux-focal-cuda12.1-py3.10-gcc9-sm80
docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks
cuda-arch-list: '8.0'
test-matrix: |
{ include: [
{ config: "inductor_huggingface_perf_cuda_a10g", shard: 1, num_shards: 3, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_huggingface_perf_cuda_a10g", shard: 2, num_shards: 3, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_huggingface_perf_cuda_a10g", shard: 3, num_shards: 3, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_timm_perf_cuda_a10g", shard: 1, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_timm_perf_cuda_a10g", shard: 2, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_timm_perf_cuda_a10g", shard: 3, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_timm_perf_cuda_a10g", shard: 4, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_timm_perf_cuda_a10g", shard: 5, num_shards: 5, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_torchbench_perf_cuda_a10g", shard: 1, num_shards: 4, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_torchbench_perf_cuda_a10g", shard: 2, num_shards: 4, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_torchbench_perf_cuda_a10g", shard: 3, num_shards: 4, runner: "linux.g5.4xlarge.nvidia.gpu" },
{ config: "inductor_torchbench_perf_cuda_a10g", shard: 4, num_shards: 4, runner: "linux.g5.4xlarge.nvidia.gpu" },
]}
selected-test-configs: ${{ inputs.benchmark_configs }}
secrets:
HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
linux-focal-cuda12_1-py3_10-gcc9-inductor-test-nightly:
name: cuda12.1-py3.10-gcc9-sm80
uses: ./.github/workflows/_linux-test.yml
needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
if: github.event.schedule == '0 7 * * *'
with:
build-environment: linux-focal-cuda12.1-py3.10-gcc9-sm80
dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-aotinductor-true-freezing_cudagraphs-true-cudagraphs_low_precision-true
docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.test-matrix }}
use-gha: anything-non-empty-to-use-gha
timeout-minutes: 720
secrets:
HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
linux-focal-cuda12_1-py3_10-gcc9-inductor-test:
name: cuda12.1-py3.10-gcc9-sm80
uses: ./.github/workflows/_linux-test.yml
needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
if: github.event_name == 'workflow_dispatch'
with:
build-environment: linux-focal-cuda12.1-py3.10-gcc9-sm80
dashboard-tag: training-${{ inputs.training }}-inference-${{ inputs.inference }}-default-${{ inputs.default }}-dynamic-${{ inputs.dynamic }}-cudagraphs-${{ inputs.cudagraphs }}-aotinductor-${{ inputs.aotinductor }}-maxautotune-${{ inputs.maxautotune }}-freezing_cudagraphs-${{ inputs.freezing_cudagraphs }}-cudagraphs_low_precision-${{ inputs.cudagraphs }}
docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-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 }}