Documentation

scale

Input

  • x:
    • Required: yes
    • Type: float32, float16
    • Description: input tensor
    • Restrictions: none
  • scale:
    • Required: yes
    • Type: float32, float16
    • Description: scale tensor
    • Restrictions: none
  • bias:
    • Required: no
    • Type: float32, float16
    • Description: bias tensor
    • Restrictions: none

Attribute

  • axis:
    • Required: no
    • Type: int
    • Description: the first axis of bottom[0] along which to apply scaling (default = 1). For example, if bottom[0] is 4D with shape 100 x 3 x 40 x 60, the output top[0] will have the same shape, and bottom[1] may have any of the following shapes (for the given value of axis):

      (axis == 0 == -4) 100; 100x3; 100x3x40; 100x3x40x60

      (axis == 1 == -3) 3; 3x40; 3x40x60

      (axis == 2 == -2) 40; 40x60

      (axis == 3 == -1) 60

      Furthermore, bottom[1] may have the empty shape (regardless of the value of axis) – a scalar multiplier.

    • Restriction: Must be in the range [–rank(x), rank(x)).
  • num_axes:
    • Required: no
    • Type: int
    • Description: the number of axes of the input covered by the scale parameter (default = 1). Ignored unless just one bottom is given. Set to –1 to cover all axes starting from axis. Set to 0 to multiply with a scalar.
    • Restriction: Must be in the range [–1, rank(x)). If axis_= axis > 0: axis, axis + rank(x), then: num_axes + axis_ <= rank(x).

Output

y:
  • Required: yes
  • Type: float32, float16
  • Description: output tensor
  • Restrictions: none