New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Tensorboard: graph is shown but no scalar data was found #2353
Comments
Do you have a call to |
Following on cgorman's comment: based on your Please take a look at the example code here |
Thanks guys. I was missing this:
|
I have a similar issue, except that I was running the example code: I can see IMAGES and GRAPHS, but no EVENTS or HISTOGRAMS, with the following command: I compiled the code from source. The following is output by adding "--debug":
And this is the output with "--inspect":
|
@kangxin That's strange. Based on the output from --inspect it looks like TensorBoard is detecting the scalar and histogram data. Can you upload the events file so I can inspect them? |
@danmane Thank you for quick response. I have attached the event files in |
Hi @kangxin, It looks suspiciously similar to #2607, where someone reported that the demo instance at https://www.tensorflow.org/tensorboard/index.html#events is not showing them events or histograms, so maybe it is a problem affecting a particular browser/OS. Please check if that url is working for you. If it is broken, then please tell me which browser and operating system you are using. If it is not broken, then there's probably something wrong with your particular TensorBoard, so please give me the following info:
|
OK, I've fixed that at head, see: 5334468 That said, we only guarantee support for Chrome and Firefox, and performance and experience are likely to be best with Chrome. |
Was just having similar problems, and double-checked multiple times that everything was set up correctly. Turned out my example ops were too "cheap" to compute, and adding a |
@dandelionmane I have similar problems when running the example code: |
verified, adding |
@liufuyang where to add writer.flush() function? |
@yvaish87 I'd would be best to add it to the step where you want to update the summary, for example, sth. like with tf.Session(graph=g) as sess:
sess.run(tf.global_variables_initializer())
# create FileWrite object that writes the logs
file_writer = tf.summary.FileWriter(logdir='logs/3', graph=g)
for i in range(5):
# fetch the summary from the graph
result, summary = sess.run([train_op, merged_summary],
feed_dict={some_value: i})
# write the summary to the log
file_writer.add_summary(summary=summary, global_step=i)
file_writer.flush() |
Thanks @rasbt ! I have the same issue, and either writer.flush or writer.close work for me: Before using flush/close:
after:
|
try to upgrade tensorflow and ensure tensboard and tensorflow are running same versions...i see these inconsitent problems if both are of different versions.. |
self.summary_writer.flush() worked for me |
I can see the graph, but no scalar is shown.
Environment info
Mac OS 10.11.4
Anaconda Python 2.7
Tensoflow installed from binary pip package
The output from python -c "import tensorflow as tf; print(tf.version)": 0.8.0
About the code
I am trying to log values of loss using:
tf.scalar_summary('loss',loss)
I have the writer set to a particular folder:
summary_writer = tf.train.SummaryWriter(LogDir, sess.graph)
After calling
sess.run([optimizer, loss, learning_rate, train_prediction], feed_dict=feed_dict)
I dosummary_writer.flush()
What have you tried?
I can see a log file called events.out.tfevents.1463159389.Matheuss-MacBook-Pro.local in the folder I set. The file is about 16Mb big.
I call
tensorboard --logdir=/Users/mviana/Desktop/tf/log/
--debug and I can verify the log dir is correct.On the browser I can see the graph corresponding to my deep neural network, but no scalar is shown.
If I call
tensorboard --inspect --logdir=/Users/mviana/Desktop/tf/log/
nothing butStarting TensorBoard 16 on port 6006 (You can navigate to http://0.0.0.0:6006)
is shown in the terminal.If I comment the line
tf.scalar_summary('loss',loss)
the log file size remains unchanged. So, it seems that only the graph is being logged.The text was updated successfully, but these errors were encountered: