Extend permitted processes during device isolation to all child proce… #163

	name: ROCm Pytorch Tests

	on:
	workflow_dispatch:
	push:
	branches: [main]
	pull_request:
	types: [opened, reopened, synchronize]

	concurrency:
	group: ${{ github.workflow }}-${{ github.event.pull_request.number \|\| github.ref }}
	cancel-in-progress: true

	jobs:
	build_image_and_run_gpu_tests:
	strategy:
	fail-fast: false
	matrix:
	image:
	[
	{ torch_rocm: rocm5.6, torch_pre_release: 0, rocm_version: 5.6.1 },
	{ torch_rocm: rocm5.7, torch_pre_release: 1, rocm_version: 5.7 },
	]

	runs-on: hf-amd-mi210-dev
	steps:
	- name: Checkout code
	uses: actions/checkout@v3

	- name: Build image
	run: docker build
	--file docker/rocm.dockerfile
	--build-arg USER_ID=$(id -u)
	--build-arg GROUP_ID=$(id -g)
	--build-arg TORCH_ROCM=${{ matrix.image.torch_rocm }}
	--build-arg TORCH_PRE_RELEASE=${{ matrix.image.torch_pre_release }}
	--build-arg ROCM_VERSION=${{ matrix.image.rocm_version }}
	--tag opt-bench-rocm:${{ matrix.image.rocm_version }}
	.

	- name: Run tests
	run: docker run
	--rm
	--net host
	--pid host
	--shm-size 64G
	--env USE_ROCM="1"
	--volume $HOME/.cache/huggingface:/home/user/.cache/huggingface
	--volume $(pwd):/workspace/optimum-benchmark
	--workdir /workspace/optimum-benchmark
	--device /dev/kfd
	--device /dev/dri/renderD128
	--device /dev/dri/renderD129
	--entrypoint /bin/bash
	opt-bench-rocm:${{ matrix.image.rocm_version }}
	-c "pip install -e .[test,peft,deepspeed] && pytest -k 'cuda and pytorch' -x"

Provide feedback