SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Better way to run AMD CI with different flavors (#26634)

Browse Source 
* Enable testing against mi250

* Change BERT to trigger tests

* Revert BERT's change

* AMD CI

* AMD CI

---------

Co-authored-by: Morgan Funtowicz <funtowiczmo@gmail.com>
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
This commit is contained in:
Yih-Dar
2023-10-16 16:24:30 +02:00
committed by GitHub
parent 3ef7134553
commit 12cc123359
4 changed files with 64 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
.github/workflows/self-push-amd.yml vendored
View File

@@ -1,21 +1,11 @@
name: Self-hosted runner AMD GPU (push)
on:
workflow_run:
workflows: ["Self-hosted runner (push-caller)"]
branches: ["main"]
types: [completed]
push:
branches:
- ci_*
- ci-*
paths:
- "src/**"
- "tests/**"
- ".github/**"
- "templates/**"
- "utils/**"
repository_dispatch:
workflow_call:
inputs:
gpu_flavor:
required: true
type: string
env:
HF_HOME: /mnt/cache
@@ -45,8 +35,7 @@ jobs:
strategy:
matrix:
machine_type: [single-gpu, multi-gpu]
gpu_flavor: [mi210]
runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ matrix.gpu_flavor }}']
runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
container:
image: huggingface/transformers-pytorch-amd-gpu-push-ci # <--- We test only for PyTorch for now
options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -65,8 +54,7 @@ jobs:
strategy:
matrix:
machine_type: [single-gpu, multi-gpu]
gpu_flavor: [mi210]
runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ matrix.gpu_flavor }}']
runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
container:
image: huggingface/transformers-pytorch-amd-gpu-push-ci # <--- We test only for PyTorch for now
options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -164,8 +152,7 @@ jobs:
matrix:
folders: ${{ fromJson(needs.setup_gpu.outputs.matrix) }}
machine_type: [single-gpu, multi-gpu]
gpu_flavor: [mi210]
runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ matrix.gpu_flavor }}']
runs-on: [self-hosted, docker-gpu, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
container:
image: huggingface/transformers-pytorch-amd-gpu-push-ci # <--- We test only for PyTorch for now
options: --device /dev/kfd --device /dev/dri --env HIP_VISIBLE_DEVICES --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -321,7 +308,7 @@ jobs:
CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_AMD }}
ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
CI_EVENT: push
CI_EVENT: Push CI (AMD) - ${{ inputs.gpu_flavor }}
CI_TITLE_PUSH: ${{ github.event.head_commit.message }}
CI_TITLE_WORKFLOW_RUN: ${{ github.event.workflow_run.head_commit.message }}
CI_SHA: ${{ env.CI_SHA }}