Update llm perf #377

	name: API ROCm 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_api_rocm_tests:
	strategy:
	fail-fast: false
	matrix:
	image: [{ rocm_version: 5.7.1, torch_rocm: rocm5.7 }]

	runs-on: [single-gpu, amd-gpu, mi250, ci]

	steps:
	- name: Checkout code
	uses: actions/checkout@v3

	- name: Target devices
	run: \|
	echo "DEVICE: $DEVICE"
	echo "DEVICE=$DEVICE" >> $GITHUB_ENV

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

	- name: Run tests
	uses: addnab/docker-run-action@v3
	env:
	DEVICE: ${{ env.DEVICE }}
	HF_TOKEN: ${{ secrets.HF_TOKEN }}
	PUSH_REPO_ID: optimum-benchmark/rocm
	with:
	image: opt-bench-rocm:${{ matrix.image.rocm_version }}
	options: \|
	--rm
	--shm-size 64G
	--env HF_TOKEN
	--env PUSH_REPO_ID
	--device /dev/kfd/
	--device /dev/dri/${{ env.DEVICE }}
	--volume ${{ github.workspace }}:/workspace
	--workdir /workspace
	run: \|
	pip install -e .[testing,timm,diffusers,codecarbon]
	pytest -s -x -k "api and cuda"

Provide feedback