Uncaught exception in projector for non-2D tensor

If the tensor to-be-visualized in the embedding projector is not 2D, there appears to be an uncaught exception (from inspecting the JavaScript Console). This took me awhile to debug (I had singleton dimensions in my tensor). A friendly error message would be helpful.

import numpy as np
import tensorflow as tf
from tensorflow.contrib.tensorboard.plugins import projector

x = np.random.rand(100, 256)
x = np.expand_dims(x, 1)  # will trigger the uncaught exception

x_var = tf.Variable(x, name='x')

config = projector.ProjectorConfig()
embedding = config.embeddings.add()
embedding.tensor_name = x_var.name

summary_writer = tf.summary.FileWriter('log')
saver = tf.train.Saver([x_var])

sess = tf.Session()
sess.run(x_var.initializer)
saver.save(sess, '{}/model.ckpt'.format('log'))
projector.visualize_embeddings(summary_writer, config)

From the JavaScript Console:

Uncaught TypeError: Cannot read property 'length' of undefined
    at (index):147378
    at Array.filter (<anonymous>)
    at (index):147376
    at XMLHttpRequest.xhr.onload ((index):143674)

Start with the embedding projector and reproduce the failure using the provided Python example, then inspect the browser JavaScript Console for the uncaught exception. Trace the projector's handling of non-2D tensors and add a friendly error message; done means the example no longer produces an uncaught exception and clearly explains the invalid tensor shape.