Add llama.cpp backend #549

	name: API ROCm Tests

	on:
	workflow_dispatch:
	push:
	branches:
	- main
	paths:
	- .github/workflows/test_api_rocm.yaml
	- "optimum_benchmark/**"
	- "docker/**"
	- "tests/**"
	- "setup.py"
	pull_request:
	branches:
	- main
	paths:
	- .github/workflows/test_api_rocm.yaml
	- "optimum_benchmark/**"
	- "docker/**"
	- "tests/**"
	- "setup.py"

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

	env:
	IMAGE: ghcr.io/huggingface/optimum-benchmark:latest-rocm

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

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

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

	- name: Unroot docker image
	run: \|
	docker build --build-arg IMAGE=${{ env.IMAGE }} --build-arg USER_ID=$(id -u) --build-arg GROUP_ID=$(id -g) -t ${{ env.IMAGE }}-unroot docker/unroot

	- 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: ${{ env.IMAGE }}-unroot
	options: \|
	--rm
	--shm-size 64G
	--env HF_TOKEN
	--env PUSH_REPO_ID
	--device /dev/kfd
	--device /dev/dri/${{ 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