Tensorboard webapp won't start

Herve
Herve Partner, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Dataiku DSS Adv Designer, Registered Posts: 58 Partner
edited July 16 in Academy Discussions

I followed instructions at https://academy.dataiku.com/image-classification-with-visual-tools-open/583580 

2021-04-24 17:25:46,735 INFO Starting Webapp backend2021-04-24 17:25:46,737 INFO Serving dependencies for html2021-04-24 17:25:46,741 INFO Serving html2021-04-24 17:25:46,742 INFO Starting backend for web app: DKU_TUTORIAL_IMAGE_CLASSIFICATION_THE_VISUAL_WAY.yADy4U0/home/hblanc/DATA_DIR/code-envs/python/PYTHON_DL/lib/python3.6/site-packages/markdown/util.py:87: DeprecationWarning: SelectableGroups dict interface is deprecated. Use select.  INSTALLED_EXTENSIONS = metadata.entry_points().get('markdown.extensions', ())2021-04-24 17:25:56,119 WARNING From /home/hblanc/DATA_DIR/plugins/installed/deeplearning-image-cpu/python-lib/tensorboard_handle.py:13: The name tf.logging.WARN is deprecated. Please use tf.compat.v1.logging.WARN instead.2021-04-24 17:25:56,119 WARNING From /home/hblanc/DATA_DIR/plugins/installed/deeplearning-image-cpu/python-lib/tensorboard_handle.py:19: The name tf.logging.set_verbosity is deprecated. Please use tf.compat.v1.logging.set_verbosity instead.2021-04-24 17:25:56,194 ERROR Backend main loop failedTraceback (most recent call last):  File "/home/hblanc/dataiku-dss-9.0.1/python/dataiku/webapps/backend.py", line 71, in serve    exec(command["code"], globals(), globals()) # in globals so that flask can find them  File "<string>", line 25, in <module>  File "/home/hblanc/DATA_DIR/plugins/installed/deeplearning-image-cpu/python-lib/tensorboard_handle.py", line 23, in __init__    app = self.__get_tb_app(logs_path)  File "/home/hblanc/DATA_DIR/plugins/installed/deeplearning-image-cpu/python-lib/tensorboard_handle.py", line 50, in __get_tb_app    assets_zip_provider=tb_default.get_assets_zip_provider(),AttributeError: module 'tensorboard.default' has no attribute 'get_assets_zip_provider'

Tagged:
· Share on FacebookShare on Twitter

Best Answer

  • Sergey
    Sergey Dataiker, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS Core Concepts Posts: 365 Dataiker
    edited July 17 Answer ✓

    Hi @Herve

    My bad, I haven't rebuilt the env so messed up with the package versions. Please make sure you have the next versions set: 

    tensorflow==1.4.0
keras==2.3.1
keras-preprocessing==1.1.0
scikit-learn>=0.20,<0.21
scipy>=1.1,<1.2
statsmodels>=0.9,<0.10
jinja2>=2.10,<2.11
flask==0.12.2
h5py==2.7.1
pillow==5.1.0

    and rebuild (check Rebuild env option) the code env. In this case, tensorflow will install tensorflow-tensorboard==0.4.0.

    · Share on FacebookShare on Twitter

Answers

  • Sergey
    Sergey Dataiker, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS Core Concepts Posts: 365 Dataiker

    Hi @Herve
    ,

    There is high chance you have installed a higher version of tensorboard which has 'get_assets_zip_provider' attribute deprecated.

    Please make sure you have followed the instructions and install the packages from the "Set of Packages". This way DSS will create a code env with precise versions of visual ML libraries.

    Thanks

    · Share on FacebookShare on Twitter
  • Herve
    Herve Partner, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Dataiku DSS Adv Designer, Registered Posts: 58 Partner

    Hi @sergeyd
    ,

    Please note I followed instructions https://academy.dataiku.com/image-classification-with-visual-tools-open which, I believe, did not mention tensorboard version.

    Which version of tensorboard should be used ? From the python code env I got

    tensorboard==1.15.0 (which is, I believe, what you get by default when creating a dev env for Visual Deep Learning)

    Thanks, Herve

    · Share on FacebookShare on Twitter
  • Sergey
    Sergey Dataiker, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS Core Concepts Posts: 365 Dataiker

    Hi @Herve

    Yes, the version is not mentioned in the academy but many users just install the packages manually without using predefined "Set of packages" hence getting different packages versions. So wanted to double-check. Yes, tensorboard should be 1.15 (same as tensorflow).

    I have followed the steps in that article before and had no issues. Let me check it again.

    · Share on FacebookShare on Twitter
  • Herve
    Herve Partner, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Dataiku DSS Adv Designer, Registered Posts: 58 Partner

    Hi @sergeyd

    I tend not to fiddle with packages versions n my own. That's why I use Dataiku : I don't like that kind of manual tedious work, it generally means trouble and I don't have much time for that

    · Share on FacebookShare on Twitter
  • Sergey
    Sergey Dataiker, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS Core Concepts Posts: 365 Dataiker
    edited July 17

    Yes, that's the best practice.

    Back to the issue: I've followed the python code and it says: 

    # To work, your web-app requires to run on a code-env with the following libraries installed:
    # tensorflow==1.4.0
    # flask==0.12.2

    So please navigate the code env you have used for this webapp and set the next versions for the packages: 

    tensorflow==1.4.0
tensorboard==1.4.0
flask==0.12.2

    I just checked and can confirm this works.

    · Share on FacebookShare on Twitter
  • Herve
    Herve Partner, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Dataiku DSS Adv Designer, Registered Posts: 58 Partner
    edited July 17

    I tried creating a code-env with 

    tensorflow==1.4.0
tensorboard==1.4.0
flask==0.12.2

    and I got the following:

    Errors were encountered

    • ERROR
      ERR_CODEENV_UPDATE_FAILED
      Environment update failed
      /home/hblanc/DATA_DIR/code-envs/python/TensorboardWebapp/bin/python failed (exit code: 1)
    Updating code environment according to spec ...Installing from Pip requirementsERROR: Could not find a version that satisfies the requirement tensorboard==1.4.0ERROR: No matching distribution found for tensorboard==1.4.0WARNING: You are using pip version 20.3.3; however, version 21.1 is available.You should consider upgrading via the '/home/hblanc/DATA_DIR/code-envs/python/TensorboardWebapp/bin/python -m pip install --upgrade pip' command.
    · Share on FacebookShare on Twitter
  • Herve
    Herve Partner, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS ML Practitioner, Dataiku DSS Core Concepts, Dataiku DSS Adv Designer, Registered Posts: 58 Partner

    Thanks @sergeyd
    , it worked.

    You might want to update https://academy.dataiku.com/image-classification-with-visual-tools-open/583580 too.

    Thanks again, Herve

    · Share on FacebookShare on Twitter
  • Sergey
    Sergey Dataiker, Dataiku DSS Core Designer, Dataiku DSS & SQL, Dataiku DSS Core Concepts Posts: 365 Dataiker

    Glad to hear this works for you.

    Yes, we already working on updating the course.

    · Share on FacebookShare on Twitter
Bookmark this topic

Categories

Setup Info
    Tags
      Help me…