UNIT2 - TensorBoard not loading

[paedusan]

Hi,

I’ve just been through UNIT 2. But when I place the address on the browser (accessing http://13.210.217.1:6006/), I get this message:

No dashboards are active for the current data set.

This is what I’ve followed:

user:~/catkin_ws/src/tf_unit1_pkg/scripts$ tensorboard --logdir=learning_logs &
[1] 13127
user:~/catkin_ws/src/tf_unit1_pkg/scripts$ public_ip
13.210.217.1
user:~/catkin_ws/src/tf_unit1_pkg/scripts$ TensorBoard 1.5.1 at http://rosdscomputer:6006 (Press CTRL+C to quit) 

What went wrong here?

[duckfrost2]

Hi. I just tried it and it seems to work::

So what I did as to follow the instructions in the notebook as so:

roscd tf_unit1_pkg; git clone Bitbucket
cd course_tflow_image_student_data/show_case_pb_models
roscd tf_unit1_pkg/scripts
python import_pb_to_tensorboard.py --model_dir=…/course_tflow_image_student_data/show_case_pb_models/classify_image_graph_def.pb --log_dir=learning_logs

And now that we have the DB model generated, then we can launch the tensorboard:

tensorboard --logdir=learning_logs

And connect to the IP that you have in RobotIgniteAcademy:

public_ip

→ WhateverIP it gives

And open in your browser the following address:

WhateverIP:6006

Please let us know if you continue with the issue and give us a detailed explanation of the steps you do to reproduce that error. Thanks

[yijionglin.bourne]

Hello,

When I put the public_id into the browser, it still does not work.

[yijionglin.bourne]

Also, according to the current courses, there is no tf_unit1_pkg/script or import_pb_to_tensorboard.py.

Could you pls help us to get the Tensorboard work?

[duckfrost2]

Hi,

Sorry , the package is my_tf_course_pkg : https://bitbucket.org/theconstructcore/tensorflow2_course_solutions/src/master/my_tf_course_pkg/

As for the tensorboard, just detected that there mustbe an issue with the port 6006. We will work on it as soon as possible to fix it.

As a workaround I found that internally you CAN access tensorboard. So here is teh workaround:

1. Open a terminal in the course and there just install firefox browser:

   sudo apt update
   sudo apt install firefox

2. Launch it and open the GraphycalTools ( The icon of ascreen beside the IDE window ). Execute this command in the shell to start firefox

   firefox

3. Write on the browser the following addres: 0.0.0.0:6006, and it should work like a charm. and look like this:

[yijionglin.bourne]

@duckfrost

Thanks sooooo muuuuch for the real quick and elaborate reply!

IT WORKS!

[duckfrost2]

YEAH! Great to hear that!

[duckfrost2]

Hi,

Just an update. Now the solution has been deployed into production , so now you don’t need this workaround.
Now there is an alias

tensorboard_address 

in webshell that gives you the exact link to which you have to connect once you launch tensorboard ;).