llm-swarm backend integration for slurm clusters #77

	name: CLI CUDA 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_cli_cuda_pytorch_tests:
	strategy:
	fail-fast: false
	matrix:
	image:
	[
	{ torch_cuda: cu118, torch_pre_release: 0, cuda_version: 11.8.0 },
	{ torch_cuda: cu121, torch_pre_release: 1, cuda_version: 12.1.1 },
	]

	runs-on: nvidia-gpu
	steps:
	- name: Checkout
	uses: actions/checkout@v3

	- name: Build image
	run: docker build
	--file docker/cuda.dockerfile
	--build-arg USER_ID=$(id -u)
	--build-arg GROUP_ID=$(id -g)
	--build-arg TORCH_CUDA=${{ matrix.image.torch_cuda }}
	--build-arg CUDA_VERSION=${{ matrix.image.cuda_version }}
	--tag opt-bench-cuda:${{ matrix.image.cuda_version }}
	.

	- name: Get GPUs with most free memory
	id: get_devices
	run: \|
	echo "::set-output name=devices::$(nvidia-smi --query-gpu=memory.free,index --format=csv,noheader,nounits \| sort -n -k1 \| tail -n 2 \| awk -F', ' '{print $2}' \| xargs echo -n \| sed 's/ /,/g' \| awk '{print $0}')"

	- name: Run tests
	run: docker run
	--rm
	--pid host
	--shm-size 64G
	--env USE_CUDA="1"
	--gpus '"device=${{ steps.get_devices.outputs.devices }}"'
	--volume $(pwd):/workspace/optimum-benchmark
	--workdir /workspace/optimum-benchmark
	--entrypoint /bin/bash
	opt-bench-cuda:${{ matrix.image.cuda_version }}
	-c "pip install -e .[testing,diffusers,timm,deepspeed,peft] && pytest -k 'cli and cuda and pytorch' -x"

Provide feedback