Skip to content

Allow multiple runs and handle connection communication errors #813

Allow multiple runs and handle connection communication errors

Allow multiple runs and handle connection communication errors #813

name: API ROCm Tests
on:
workflow_dispatch:
push:
branches:
- main
pull_request:
branches:
- main
types:
- opened
- reopened
- synchronize
- labeled
- unlabeled
concurrency:
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
jobs:
run_api_rocm_tests:
if: ${{
(github.event_name == 'push') ||
(github.event_name == 'workflow_dispatch') ||
contains( github.event.pull_request.labels.*.name, 'api') ||
contains( github.event.pull_request.labels.*.name, 'rocm') ||
contains( github.event.pull_request.labels.*.name, 'api_rocm')
}}
runs-on: [self-hosted, amd-gpu, single-gpu]
container:
image: ghcr.io/huggingface/optimum-benchmark:latest-rocm
options: --ipc host
--shm-size "16gb"
--group-add video
--device /dev/kfd
--device /dev/dri
--env ROCR_VISIBLE_DEVICES
--env HIP_VISIBLE_DEVICES=0
steps:
- name: Checkout
uses: actions/checkout@v4
- name: Install dependencies
run: |
pip install -e .[testing,timm,diffusers,codecarbon]
- name: Run tests
env:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
PUSH_REPO_ID: optimum-benchmark/rocm
run: |
pytest -s -x -k "api and cuda"