CODEWITHSUNDEEP
tensorflowcallbackgoogle-colaboratorytensorboard
cLwill
cLwill

Reputation: 102

import import tensorflow_docs error on Colab

I've been using colab for deep learning for a over a month and all of a sudden import tensorflow_docs as tfdocs stopped working. Is anyone encountering the same issues???

I'm running tf.version 2.8.2

> !pip install git+https://github.com/tensorflow/docs
> 
> import tensorflow_docs as tfdocs   
> import tensorflow_docs.modeling   
> import tensorflow_docs.plots

Error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-10-8469b6568dc0> in <module>()
      2 get_ipython().system('pip install git+https://github.com/tensorflow/docs')
      3 
----> 4 import tensorflow_docs as tfdocs
      5 import tensorflow_docs.modeling
      6 import tensorflow_docs.plots

4 frames
/usr/local/lib/python3.7/dist-packages/tensorflow_docs/api_generator/reference_resolver.py in ReferenceResolver()
     86       py_module_names: Union[list[str], dict[str, str]],
     87       link_prefix: Optional[str] = None,
---> 88       physical_path: Optional[dict[str, str]] = None,
     89   ):
     90     """Initializes a Reference Resolver.

TypeError: 'type' object is not subscriptable

UPDATE 01-Sep-2022

I updated python in colab to 3.9.1 using @Kor suggestion and confirmed it. Then ran:

> !pip install git+https://github.com/tensorflow/docs
> 
> import tensorflow_docs as tfdocs

The tensorflow doc completed installation but still threw an error on import:

ModuleNotFoundError                       Traceback (most recent call last)
    <ipython-input-15-2bc3446903cd> in <module>
    ----> 1 import tensorflow_docs as tfdocs
          2 import tensorflow_docs.modeling
          3 import tensorflow_docs.plots
          4 
          5 print("Version: ", tf.__version__)

    ModuleNotFoundError: No module named 'tensorflow_docs'

Upvotes: 4

Views: 2379

Answers (6)

ngbinhming
ngbinhming

Reputation: 1

After

!pip install git+https://github.com/tensorflow/docs

Open (if you are using Colab)

/usr/local/lib/python3.7/dist-packages/tensorflow_docs/api_generator/reference_resolver.py

and add from __future__ import annotations at the BEGINNING OF THE FILE and then import

import tensorflow_docs as tfdocs
import tensorflow_docs.modeling
import tensorflow_docs.plots

Source: https://github.com/tensorflow/docs/pull/2116/files

Upvotes: 0

Kor
Kor

Reputation: 48

I assuming you are using Google Colab. As I struggled the same with Google Colab since the default Google Colab is in Python 3.7. Here's what I did to make Google Colab upgrade to Python 3.9

!wget -O mini.sh https://repo.anaconda.com/miniconda/Miniconda3-py39_4.9.2-Linux-x86_64.sh
!chmod +x mini.sh
!bash ./mini.sh -b -f -p /usr/local
!conda install -q -y jupyter
!conda install -q -y google-colab -c conda-forge
!python -m ipykernel install --name "py39" --user

Hope this helps you to move on as I did!

Upvotes: 1

Mark McDonald
Mark McDonald

Reputation: 8180

---> 88       physical_path: Optional[dict[str, str]] = None,
     89   ):
     90     """Initializes a Reference Resolver.

TypeError: 'type' object is not subscriptable

This error is caused by the dict[str, str] line. This is PEP-585 compliant code that works in Python 3.9, but it looks like Colab is still running Python 3.7, so it's failing.

Feel free to send us PRs adding from __future__ import annotations to the affected files if you spot these, but ultimately we run 3.9 internally at Google so these may keep sneaking through.

Some other ideas that you might be able to make work:

  • Use a Python 3.9 kernel for colab (e.g. via Google Cloud)
  • Strip out type hints with something like strip-hints

Upvotes: 1

Mohammad Javad Ahmadi
Mohammad Javad Ahmadi

Reputation: 66

You can try this:

!pip install -q git+https://github.com/MJAHMADEE/docs
import tensorflow_docs as tfdocs   
import tensorflow_docs.modeling   
import tensorflow_docs.plots

Upvotes: 2

P.cordo
P.cordo

Reputation: 31

I had also a problem with a script using tensor flow in colab, that had been running for months without any problem. Suddenly I start having the same error as you mention:

TypeError: 'type' object is not subscriptable

The script finished in the following line, at the top of the script:

from tensorflow_docs.vis import embed

I have just commented the line, and the script started working again... :

# from tensorflow_docs.vis import embed

I hope this may help some of you having the same error.

Upvotes: 0

ClaudiaR
ClaudiaR

Reputation: 3414

I have this problem too, but sadly I don't have a solution. However, this is what I've tried so far:

  • Downgrading to Tensorflow 2.6
  • Installing a previous version of tensorflow_docs with !pip install git+https://github.com/tensorflow/docs@my_version choosing as my_version a commit from a few weeks ago.
  • Installing tensorflow_docs in a different way, first downloading it with !git clone https://github.com/tensorflow/docs and then installing it via !pip install docs/

All without success.

Anyway from the error it seems it is a problem of the tensorflow_docs'ReferenceResolver, it looks like it is not able to find the library. However with the following instruction tensorflow-docs is found:

!pip3 freeze | grep tensorflow

Upvotes: 1

Related Questions