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_array_to_buffer

float32_array_to_buffer()

max.driver.float32_array_to_buffer(array, *, dtype, device)

source

Creates a Buffer of dtype on device from float32 data.

Converts on the host, so no cast graph is compiled. For DType.bfloat16, rounds with float32_to_bfloat16_as_uint16(); for other dtypes, casts with NumPy.

Parameters:

• array (ndarray[tuple[Any, ...], dtype[float32]]) – The float32 host data.
• dtype (DType) – The element type of the returned buffer.
• device (Device) – The device to place the returned buffer on.

Returns:

A buffer with the same shape as array.

Return type:

Buffer