Skip to main content

Module dequant_to_f16

mlx_native::ops

Module dequant_to_f16 

Source

Functionsยง

dispatch_dequant_to_f16
Dispatch the whole-tensor dequant-to-F16 kernel.
materialize_f16_shadow
One-shot helper: allocate an F16 shadow buffer + dispatch + commit-and-wait.