3 ms·IEEE 16-bit float is supported on the M1 and is generally all you need for AI inference.by jacobgorm 3y agoIEEE 16-bit float is supported on the M1 and is generally all you need for AI inference.