triton.language.maximum

triton.language.maximum(x, y, propagate_nan: ~triton.language.core.constexpr = <PROPAGATE_NAN.NONE: 0>)

Computes the element-wise maximum of x and y.

Parameters:
  • x (Block) – the first input tensor

  • y (Block) – the second input tensor

  • propagate_nan (tl.PropagateNan) – whether to propagate NaN values.

See also

tl.PropagateNan