Add the cuda dequantize f16 kernels. (#2137) #5272

	on:
	push:
	branches:
	- main
	pull_request:

	name: Continuous integration

	jobs:
	check:
	name: Check
	runs-on: ${{ matrix.os }}
	strategy:
	matrix:
	os: [ubuntu-latest, windows-latest, macOS-latest]
	rust: [stable]
	steps:
	- uses: actions/checkout@v2
	- uses: actions-rs/toolchain@v1
	with:
	profile: minimal
	toolchain: ${{ matrix.rust }}
	override: true
	- uses: actions-rs/cargo@v1
	with:
	command: check
	args: --workspace

	test:
	name: Test Suite
	runs-on: ${{ matrix.os }}
	strategy:
	matrix:
	os: [ubuntu-latest, windows-latest, macOS-latest]
	rust: [stable]
	steps:
	- uses: actions/checkout@v2
	- uses: actions-rs/toolchain@v1
	with:
	profile: minimal
	toolchain: ${{ matrix.rust }}
	override: true
	- uses: actions-rs/cargo@v1
	with:
	command: test
	args: --workspace

	fmt:
	name: Rustfmt
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@v2
	- uses: actions-rs/toolchain@v1
	with:
	profile: minimal
	toolchain: stable
	override: true
	- run: rustup component add rustfmt
	- uses: actions-rs/cargo@v1
	with:
	command: fmt
	args: --all -- --check

	clippy:
	name: Clippy
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@v2
	- uses: actions-rs/toolchain@v1
	with:
	profile: minimal
	toolchain: stable
	override: true
	- run: rustup component add clippy
	- uses: actions-rs/cargo@v1
	with:
	command: clippy
	args: --workspace --tests --examples -- -D warnings

Provide feedback