A high-performance inference engine for AI models
"""Launch the MLX benchmark adapter in its own uv project."""
import pytest
from .engine_tests import EngineTests
MODEL = "mlx-community/Qwen3.5-0.8B-4bit"
def add_options(parser: pytest.Parser) -> None:
group = parser.getgroup("MLX")
group.addoption("--mlx-model", default=MODEL, help="Target model local path or HuggingFace id.")
class TestMLX(EngineTests):
engine_name = "mlx"
@staticmethod
def command(config: pytest.Config) -> list[str]:
return ["uv", "run", "--project", "engine-mlx", "bench-mlx", "--model", config.getoption("mlx_model")]