Operator definition: IF

[dmatveev]

IF is an example of a quite basic high-order construct.

• OpenVINO

  • if_op, v8::If
    • Semantics:
      • Takes three Outputs in constructor (condition, true-value, false-value).
      • Doesn't look like a nested, scoped construct (code remains "flat" for the user)
    • Example: n/a

• PyTorch

  • where
    • Semantics:
      • Eager/Pixelwise, no subgraph propagation??
    • Example:

      >>> x = torch.randn(3, 2)
      >>> y = torch.ones(3, 2)
      >>> x
      tensor([[-0.4620,  0.3139],
              [ 0.3898, -0.7197],
              [ 0.0478, -0.1657]])
      >>> torch.where(x > 0, 1.0, 0.0)
      tensor([[0., 1.],
              [1., 0.],
              [1., 0.]])
      >>> torch.where(x > 0, x, y)
      tensor([[ 1.0000,  0.3139],
              [ 0.3898,  1.0000],
              [ 0.0478,  1.0000]])
      >>> x = torch.randn(2, 2, dtype=torch.double)
      >>> x
      tensor([[ 1.0779,  0.0383],
              [-0.8785, -1.1089]], dtype=torch.float64)
      >>> torch.where(x > 0, x, 0.)
      tensor([[1.0779, 0.0383],
            [0.0000, 0.0000]], dtype=torch.float64)

• TensorFlow

  • cond
    • Semantics
      • Takes functions as true/false branches. Predicate is tensor value (lazy evaluation?)
      • Standard Python's if statement is also converted to cond per documentation (maybe it is the case in PyTorch too).
    • Example:

      @tf.function
      def fun1(x,y):
        if x > 0:  # AutoGraph converts if-statement to tf.cond().
          z = y+1
        else:
          z = y-1
        return z
      fun1(tf.constant(7), tf.constant(3)).numpy()
      4