maximum¶

nvtripy.maximum(lhs: Tensor | Number, rhs: Tensor | Number) → Tensor[source]¶

Performs an elementwise maximum.

Parameters:

lhs (Tensor | Number) – The first input tensor.
rhs (Tensor | Number) – The second input tensor. It should be broadcast-compatible.

Returns:

A new tensor with the broadcasted shape.

Return type:

Tensor

INPUT REQUIREMENTS:

lhs.dtype is one of [float32, float16, bfloat16, int8, int32, int64, bool], and
rhs.dtype == lhs.dtype

OUTPUT GUARANTEES:

return[0].dtype == lhs.dtype

Example

a = tp.Tensor([1.0, 6.0])
b = tp.Tensor([2.0, 3.0])
output = tp.maximum(a, b)

Local Variables¶

>>> a
tensor([1, 6], dtype=float32, loc=cpu:0, shape=(2,))

>>> b
tensor([2, 3], dtype=float32, loc=cpu:0, shape=(2,))

>>> output
tensor([2, 6], dtype=float32, loc=gpu:0, shape=(2,))