DirectML

Microsoft open low-level ML API — GPU-accelerated inference on any DirectX 12 device.