ServingRuntime: add GPU configuration #8

	name: Build and Push

	on:
	push:
	branches: ["main"]
	paths-ignore:
	- "**.md"
	pull_request:
	paths-ignore:
	- "**.md"
	workflow_dispatch:

	env:
	QUAY_REPO: "quay.io/dtrifiro/llama-cpp-python-serving"
	CI: true
	DOCKER_BUILDKIT: 1
	FORCE_COLOR: 1

	jobs:
	build-and-push:
	name: Build and Push Image
	runs-on: ubuntu-latest

	steps:
	- uses: actions/checkout@v3
	- name: "Setup Docker Buildx"
	uses: docker/setup-buildx-action@v2
	- name: "Build"
	run: \|
	date="$(date +%Y%m%d%H%M)"
	rev="$(git rev-parse --short HEAD)"

	docker build -t "${QUAY_REPO}:latest" -t "${QUAY_REPO}:${rev}" -t "${QUAY_REPO}:${date}" .

	- name: "Build CUDA image"
	run: \|
	date="$(date +%Y%m%d%H%M)"
	rev="$(git rev-parse --short HEAD)"

	docker build -t "${QUAY_REPO}:latest-cuda" -t "${QUAY_REPO}:${rev}-cuda" -t "${QUAY_REPO}:${date}-cuda" -f Dockerfile.cuda .

	- name: "Push to quay"
	if: github.ref == 'refs/heads/main'
	run: \|
	docker login -u="${{secrets.QUAY_USERNAME}}" -p="${{secrets.QUAY_TOKEN}}" quay.io

	docker push --all-tags "${QUAY_REPO}"

Provide feedback