Support complex type in onesLike

[Lewuathe]

FEATURE

As TensorFlow supports complex type in ones_like ops, we can support complex64 tensor in onesLike ops.

See: https://www.tensorflow.org/api_docs/python/tf/ones_like

>>> tf.ones_like([[1,2,3], [1,2,3]], dtype='complex64').eval()
array([[1.+0.j, 1.+0.j, 1.+0.j],
       [1.+0.j, 1.+0.j, 1.+0.j]], dtype=complex64)

---

This change is 

[nsthorat]

shouldn't these all be ones?

[Lewuathe]

ones_like of TensorFlow core generates a complex number with real 1 and imaginary 0. So I think it's consistent with TensorFlow core implementation.

>>> tf.ones_like([1,2,3], dtype='complex64')
<tf.Tensor: id=18, shape=(3,), dtype=complex64, numpy=array([1.+0.j, 1.+0.j, 1.+0.j], dtype=complex64)>

[nsthorat]

Reviewed 2 of 2 files at r1.
Reviewable status: complete! 1 of 1 approvals obtained (waiting on @Lewuathe)

---

src/ops/array_ops_test.ts, line 521 at r1 (raw file):

Previously, Lewuathe (Kai Sasaki) wrote…

ones_like of TensorFlow core generates a complex number with real 1 and imaginary 0. So I think it's consistent with TensorFlow core implementation.

>>> tf.ones_like([1,2,3], dtype='complex64')
<tf.Tensor: id=18, shape=(3,), dtype=complex64, numpy=array([1.+0.j, 1.+0.j, 1.+0.j], dtype=complex64)>

Ah -- apologies that's correct!