use max auto-tune no cuda graphs for max perf #63

	name: Tests on GPU

	on:
	push:
	branches:
	- main
	pull_request:
	# This will trigger the workflow for pull requests to any branch
	types: [opened, synchronize, reopened]

	jobs:
	gpu-tests:
	name: python
	runs-on: self-hosted

	steps:
	- uses: actions/checkout@v4
	with:
	submodules: true

	- name: Install uv
	uses: astral-sh/setup-uv@v3
	with:
	enable-cache: true
	cache-dependency-glob: "uv.lock"

	- name: Set up Python
	run: uv python install 3.10.13

	- name: Install the project
	run: uv sync --all-extras --dev
	- name: Run tests
	run: uv run pytest tests

Provide feedback