Optimum TPU / Test TGI on TPU (slow tests) #203

	name: Optimum TPU / Test TGI on TPU (slow tests)

	on:
	# This can be used to automatically publish nightlies at UTC nighttime
	schedule:
	- cron: '0 2 * * *' # run at 2 AM UTC
	# This can be used to allow manually triggering nightlies from the web interface
	workflow_dispatch:

	concurrency:
	group: ${{ github.workflow }}-${{ github.head_ref \|\| github.run_id }}
	cancel-in-progress: true

	jobs:
	do-the-job:
	name: Build and Run slow tests
	runs-on:
	group: gcp-ct5lp-hightpu-8t
	container:
	image: us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla:r2.4.0_3.10_tpuvm
	options: --shm-size "16gb" --ipc host --privileged ${{ vars.V5_LITEPOD_8_ENV}} -v /mnt/hf_cache:/mnt/hf_cache
	env:
	PJRT_DEVICE: TPU
	HF_TOKEN: ${{ secrets.HF_TOKEN_OPTIMUM_TPU_CI }}
	HF_HUB_CACHE: /mnt/hf_cache/cache_huggingface
	steps:
	- name: Checkout
	uses: actions/checkout@v4

	- name: Build and test Optimum TPU (also slow tests)
	run: \|
	python -m pip install build
	make build_dist test_installs
	python -m pytest --runslow -sv tests

	- name: Build and test TGI (also slow tests)
	run: \|
	make tgi_server test_installs
	find text-generation-inference/ -name "text_generation_server-*whl" -exec python -m pip install {} \;
	python -m pytest --runslow -sv text-generation-inference/tests

Provide feedback