System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux 5.2.1-arch1-1-ARCH x86_64
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): v1.14.0-rc1-22-gaf24dc91b5
• Python version: 3.6.8
• CUDA/cuDNN version: N/A
• GPU model and memory: N/A

Describe the current behavior
Calling tf.gradients with unconnected_gradients="zero" returns scalars for unconnected resource variables.

Describe the expected behavior
Calling tf.gradients with unconnected_gradients="zero" returns appropriately shaped zero tensors for unconnected resource variables.

Code to reproduce the issue

a = tf.Variable(initial_value=[2., 3.])
b = tf.Variable(initial_value=[3., 4.], use_resource=True)
c = tf.constant(0.)
print(tf.gradients(c, [a, b], unconnected_gradients="zero"))
# => [<tf.Tensor 'zeros_like:0' shape=(2,) dtype=float32>,
#     <tf.Tensor 'zeros_like_1:0' shape=() dtype=float32>]