Tensor Core enabled for mixed precision on GPUs that allow it #1399

	name: CI
	on:
	push:
	branches:
	- main
	pull_request:
	branches:
	- main
	schedule:
	- cron: '0 3 * * 1'

	jobs:
	build:

	runs-on: ubuntu-20.04
	strategy:
	matrix:
	python: [3.7, 3.8]

	steps:
	- name: Setup Python
	uses: actions/setup-python@v2
	with:
	python-version: ${{ matrix.python }}
	- name: Checkout GCNN
	uses: actions/checkout@v2.2.0
	- name: Install MPI
	run: \|
	sudo apt-get update
	sudo apt-get install -y libopenmpi-dev
	sudo apt-get clean
	- name: Cache pip
	uses: actions/cache@v2
	id: cache
	with:
	path: ${{ env.pythonLocation }}
	key: ${{ env.pythonLocation }}-${{ hashFiles('requirements.txt') }}-${{ hashFiles('requirements-dev.txt') }}-${{ hashFiles('requirements-torchdep.txt') }}
	- name: Install dependencies
	if: steps.cache.outputs.cache-hit != 'true'
	run: \|
	python -m pip install --upgrade pip
	python -m pip install --upgrade -r requirements.txt -r requirements-dev.txt
	python -m pip install --upgrade -r requirements-torchdep.txt
	- name: Format black
	run: \|
	black .
	git diff --exit-code
	- name: Run pytest
	run: \|
	# FIXME: install
	export PYTHONPATH=$PYTHONPATH:~/HydraGNN
	pip list
	python -m pytest -rfEP -W error -W ignore::DeprecationWarning -W ignore::UserWarning --tb=native
	mpirun -n 2 --oversubscribe python -m pytest -rfEP -W error -W ignore::DeprecationWarning -W ignore::UserWarning --tb=native --with-mpi

Provide feedback