asc2.broadcast_to

asc2.broadcast_to(input: LocalTensor, *shape: int) → LocalTensor

Creates new tensor of a given shape broadcasting data from the input tensor.

The supported data types are: int8, int16, int32, int64, float16, bfloat16, float32.

Parameters:

  • input – The input tensor

  • shape – The target shape (can be passed as separate integers or as an iterable, e.g. list or tuple)

Returns:

A new tensor with the broadcasted shape

Return type:

LocalTensor

Raises:

  • TypeError – If input is not a LocalTensor or shape contains non-integer values

  • RuntimeError – If the input tensor shape cannot be broadcasted to the target one or shape values are not positive

Examples

Broadcast tensor to the provided shape:

input = asc2.copy_in(x, [0], [256])
result = input.broadcast_to([16,256])

The code above may act as the following:

input:   [0,1,2,3,4, ... 255]
result:  [[0,1,2,...255], [0,1,2,...255] ... [0,1,2,..255]]

This function can also be called as a member function on LocalTensor, as input.broadcast_to(...) instead of broadcast_to(input, ...).