For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Python function

float32_to_bfloat16_as_uint16

float32_to_bfloat16_as_uint16()

max.driver.float32_to_bfloat16_as_uint16(arr)

source

Converts a float32 array to bfloat16 bit patterns stored as uint16.

Rounds to nearest, ties to even, the same as a compiled float32 to bfloat16 cast. A NaN input becomes the canonical quiet NaN with the input’s sign.

Parameters:

arr (ndarray[tuple[Any, ...], dtype[float32]]) – A float32 array.

Returns:

A uint16 array of bfloat16 bit patterns with the same shape as arr.

Return type:

ndarray[tuple[Any, …], dtype[uint16]]