mirror of
https://github.com/pytorch/pytorch.git
synced 2025-10-20 21:14:14 +08:00
[CI] Switch ROCm MI300 GitHub Actions workflows from 2-GPU to 1-GPU runners (#158882)
Updated .github/actionlint.yaml to replace linux.rocm.gpu.mi300.2 with linux.rocm.gpu.mi300.1 in the supported runner list Modified all affected workflows (inductor-perf-test-nightly-rocm.yml, inductor-periodic.yml, inductor-rocm-mi300.yml, and rocm-mi300.yml) to run jobs on 1-GPU MI300 runners instead of 2-GPU runners This should help increase available runners even with same number of CI nodes. Pull Request resolved: https://github.com/pytorch/pytorch/pull/158882 Approved by: https://github.com/jeffdaily Co-authored-by: Jeff Daily <jeff.daily@amd.com>
This commit is contained in:
deedongala
committed by
PyTorch MergeBot
PyTorch MergeBot
parent
2e4e5ab4be
commit
5737372862
1
.github/actionlint.yaml
vendored
1
.github/actionlint.yaml
vendored
@ -54,6 +54,7 @@ self-hosted-runner:
|
||||
- linux.rocm.gpu.2
|
||||
- linux.rocm.gpu.4
|
||||
# gfx942 runners
|
||||
- linux.rocm.gpu.gfx942.1
|
||||
- linux.rocm.gpu.gfx942.2
|
||||
- linux.rocm.gpu.gfx942.4
|
||||
- rocm-docker
|
||||
|
||||
5
.github/actions/setup-rocm/action.yml
vendored
5
.github/actions/setup-rocm/action.yml
vendored
@ -59,11 +59,6 @@ runs:
|
||||
echo "$msg"
|
||||
exit 1
|
||||
fi
|
||||
if [[ $ngpu -eq 1 ]]; then
|
||||
echo "Error: only 1 GPU detected, at least 2 GPUs are needed for distributed jobs"
|
||||
echo "$msg"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
- name: Runner diskspace health check
|
||||
uses: pytorch/pytorch/.github/actions/diskspace-cleanup@main
|
||||
|
||||
10
.github/workflows/_rocm-test.yml
vendored
10
.github/workflows/_rocm-test.yml
vendored
@ -88,6 +88,16 @@ jobs:
|
||||
- name: Setup ROCm
|
||||
uses: ./.github/actions/setup-rocm
|
||||
|
||||
- name: Runner check GPU count (distributed jobs)
|
||||
if: ${{ contains(matrix.config, 'distributed') }}
|
||||
shell: bash
|
||||
run: |
|
||||
ngpu=$(rocminfo | grep -c -E 'Name:.*\sgfx')
|
||||
if [[ $ngpu -lt 4 ]]; then
|
||||
echo "Error: only $ngpu GPU(s) detected, at least 4 GPUs are needed for distributed jobs"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
- name: configure aws credentials
|
||||
id: aws_creds
|
||||
uses: aws-actions/configure-aws-credentials@ececac1a45f3b08a01d2dd070d28d111c5fe6722 # v4.1.0
|
||||
|
||||
@ -88,23 +88,23 @@ jobs:
|
||||
docker-image-name: ci-image:pytorch-linux-jammy-rocm-n-py3-benchmarks
|
||||
test-matrix: |
|
||||
{ include: [
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 1, num_shards: 4, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 2, num_shards: 4, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 3, num_shards: 4, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 4, num_shards: 4, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 1, num_shards: 5, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 2, num_shards: 5, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 3, num_shards: 5, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 4, num_shards: 5, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 5, num_shards: 5, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 1, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 2, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 3, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 4, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 5, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 6, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 7, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 8, num_shards: 8, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 1, num_shards: 4, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 2, num_shards: 4, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 3, num_shards: 4, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_huggingface_perf_rocm", shard: 4, num_shards: 4, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 1, num_shards: 5, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 2, num_shards: 5, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 3, num_shards: 5, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 4, num_shards: 5, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_timm_perf_rocm", shard: 5, num_shards: 5, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 1, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 2, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 3, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 4, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 5, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 6, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 7, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor_torchbench_perf_rocm", shard: 8, num_shards: 8, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
]}
|
||||
secrets: inherit
|
||||
|
||||
|
||||
30
.github/workflows/inductor-periodic.yml
vendored
30
.github/workflows/inductor-periodic.yml
vendored
@ -81,21 +81,21 @@ jobs:
|
||||
sync-tag: rocm-build
|
||||
test-matrix: |
|
||||
{ include: [
|
||||
{ config: "dynamo_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamo_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamo_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamo_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamo_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "aot_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "aot_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "aot_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "aot_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "aot_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamic_aot_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamic_aot_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamic_aot_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamic_aot_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamic_aot_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "dynamo_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamo_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamo_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamo_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamo_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "aot_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "aot_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "aot_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "aot_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "aot_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamic_aot_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamic_aot_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamic_aot_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamic_aot_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "dynamic_aot_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
]}
|
||||
secrets: inherit
|
||||
|
||||
|
||||
4
.github/workflows/inductor-rocm-mi300.yml
vendored
4
.github/workflows/inductor-rocm-mi300.yml
vendored
@ -47,8 +47,8 @@ jobs:
|
||||
docker-image-name: ci-image:pytorch-linux-jammy-rocm-n-py3
|
||||
test-matrix: |
|
||||
{ include: [
|
||||
{ config: "inductor", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "inductor", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "inductor", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
]}
|
||||
secrets: inherit
|
||||
|
||||
|
||||
12
.github/workflows/rocm-mi300.yml
vendored
12
.github/workflows/rocm-mi300.yml
vendored
@ -48,12 +48,12 @@ jobs:
|
||||
sync-tag: rocm-build
|
||||
test-matrix: |
|
||||
{ include: [
|
||||
{ config: "default", shard: 1, num_shards: 6, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "default", shard: 2, num_shards: 6, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "default", shard: 3, num_shards: 6, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "default", shard: 4, num_shards: 6, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "default", shard: 5, num_shards: 6, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "default", shard: 6, num_shards: 6, runner: "linux.rocm.gpu.gfx942.2" },
|
||||
{ config: "default", shard: 1, num_shards: 6, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "default", shard: 2, num_shards: 6, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "default", shard: 3, num_shards: 6, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "default", shard: 4, num_shards: 6, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "default", shard: 5, num_shards: 6, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
{ config: "default", shard: 6, num_shards: 6, runner: "linux.rocm.gpu.gfx942.1" },
|
||||
]}
|
||||
secrets: inherit
|
||||
|
||||
|
||||
Reference in New Issue
Block a user