Embedding projector hangs in version 1.11.0

[persiyanov]

TensorBoard embedding projector hangs on a title "Computing PCA....". I solved this problem when downgraded both TF & TensorBoard to version 1.7.0.

• TensorBoard & TensorFlow version is 1.11.0
• OS Platform and version Linux Ubuntu 16.04
• Python version 3.5

The minimal example to reproduce the issue:

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

LOG_DIR = 'logs'
embeddings = np.array([[ 0.00403953, -0.00622064, -0.01170987,  0.02049868, -0.01784528,
        -0.00844114,  0.00360506, -0.09942514,  0.032083  , -0.0056223 ],
       [-0.04307625, -0.01297725,  0.03254195,  0.00583903, -0.00986445,
        -0.02183709,  0.01153475, -0.10950625, -0.03583045, -0.03795925],
       [-0.03899275, -0.0046966 ,  0.01469164,  0.01740974,  0.01288737,
        -0.0306037 ,  0.0353404 , -0.09731412,  0.04237695,  0.02322971],
       [-0.00181839,  0.00105444, -0.01013809,  0.04349392, -0.02176337,
        -0.04931752, -0.02105442, -0.07082064,  0.00941054, -0.02279854],
       [ 0.0296278 ,  0.0406456 , -0.01181266,  0.03085812, -0.0182738 ,
        -0.01673431,  0.0178362 , -0.1398714 ,  0.0270176 , -0.00516446],
       [-0.0595875 , -0.047008  ,  0.002867  ,  0.00726775, -0.002714  ,
        -0.029009  ,  0.002205  , -0.10231575, -0.0291282 ,  0.01943275],
       [-0.00291987, -0.069231  , -0.02141933, -0.01235343,  0.000823  ,
        -0.05252166, -0.000757  , -0.11705633,  0.0241631 ,  0.02844123],
       [ 0.0141408 , -0.02469571, -0.01984159, -0.00944844,  0.00886971,
        -0.0474157 , -0.01649545, -0.13802414, -0.02138517,  0.01678829],
       [ 0.008873  ,  0.05906433, -0.020482  , -0.01096687, -0.022655  ,
        -0.04205434,  0.02315167, -0.12080999, -0.029119  , -0.03175637],
       [ 0.028179  , -0.04476625, -0.017764  ,  0.02365533,  0.00857486,
        -0.0018825 ,  0.0024118 , -0.11940325,  0.0164014 , -0.00140572]],
      dtype=np.float32)

emb_var = tf.Variable(embeddings, name='embeddings')

with tf.Session() as sess:
    saver = tf.train.Saver([emb_var])

    sess.run(emb_var.initializer)
    saver.save(sess, os.path.join(LOG_DIR, 'emb.ckpt'))

    config = projector.ProjectorConfig()
    # One can add multiple embeddings.
    embedding = config.embeddings.add()
    embedding.tensor_name = emb_var.name
    # Saves a config file that TensorBoard will read during startup.
    projector.visualize_embeddings(tf.summary.FileWriter(LOG_DIR), config)

[stephanwlee]

Thanks for the issue with the sample code! I was able to reproduce this and I have a fix ready 😃

[benelot]

Just came across this problem while replicating some results in a project that had 1.11.0 in its requirements. I can confirm that the problem exists in version 1.11.0 only and is fixed after that in version 1.12.2.

[jvence]

Yes I can confirm that this problem exists in 1.11.0