Fix: EXX suports restart scf from HexxR.csr produced by `out_chg` #3930

	name: CUDA Test

	on:
	workflow_dispatch:
	pull_request:

	concurrency:
	group: ${{ github.workflow }}-${{ github.ref }}
	cancel-in-progress: true

	jobs:
	test:
	name: Test on CUDA Build
	runs-on: nvidia
	if: github.repository_owner == 'deepmodeling'
	container:
	image: ghcr.io/deepmodeling/abacus-cuda
	volumes:
	- /tmp/ccache:/github/home/.ccache
	options: --gpus all

	steps:
	- name: Checkout
	uses: actions/checkout@v4
	with:
	submodules: recursive

	- name: Install Ccache
	run: \|
	sudo apt-get update
	sudo apt-get install -y ccache

	- name: Build
	run: \|
	nvidia-smi
	cmake -B build -DUSE_CUDA=ON -DBUILD_TESTING=ON
	cmake --build build -j4
	cmake --install build

	- name: Test
	run: \|
	cd tests/integrate
	bash Autotest.sh -n 2 -f CASES_GPU.txt

Provide feedback