Cannot import name 'gfile' from 'tensorflow'

[themolco]

I am trying to run the lucid notebook on my machine but it barfs on lucid.modelzoo.vision_models. After a long stream of files it eventually informs me

ImportError: cannot import name 'gfile' from 'tensorflow' (C:\Users\BLEH\.conda\envs\fastai\lib\site-packages\tensorflow\__init__.py)

The file that initiates the call is vision_base.py. I am using the most recent stable versions of tensorflow and lucid.

[mettinger]

Same problem here. I think the problem is that gfile was moved to tf.io.gfile in tensorflow 2.0.

[colah]

Lucid does not support TF 2.0 at present.

[Dcrawf]

Has anyone managed to find any solutions to this error?

[mariajmolina]

I was able to bypass the error with gfile using: import tensorflow.compat.v1.gfile as gfile There are also issues with tensorflow.contrib and am not sure how to fix that yet since it seems more involved.

[AwesomeLemon]

The Colab notebook linked from the Feature Visualization Distill article also fails now with this error. Easy fix is to add

!pip install tensorflow-gpu==1.15

[MariaRigaki]

In Colab you can also use the directive %tensorflow_version 1.x before importing tensorflow.

[CTAYLOR888]

I found I needed to completely uninstall Tensorflow, then all worked fine: !pip uninstall tensorflow -y !pip uninstall tensorflow-gpu -y !pip install tensorflow==1.15 !pip install tensorflow-gpu==1.15 !pip install lucid

[TheLBL]

I also try to get the notebook of this Feature Visualization thing to run, but when I use TensorFlow 1.15.0 to bypass the gfile-problem I get this: kapre 0.3.7 requires TensorFlow>=2.0.0, but you have TensorFlow 1.15.0 which is incompatible. Does anyone have an idea on what to do?

[sgbaird]

Full stack trace:

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
[<ipython-input-3-fc76b967f420>](https://localhost:8080/#) in <module>()
     12 
     13 # General lucid code
---> 14 from lucid.misc.io import save, show, load
     15 import lucid.modelzoo.vision_models as models
     16 

3 frames
[/usr/local/lib/python3.7/dist-packages/lucid/misc/io/__init__.py](https://localhost:8080/#) in <module>()
      1 from lucid.misc.io.showing import show
----> 2 from lucid.misc.io.loading import load
      3 from lucid.misc.io.saving import save, CaptureSaveContext, batch_save
      4 from lucid.misc.io.scoping import io_scope, scope_url

[/usr/local/lib/python3.7/dist-packages/lucid/misc/io/loading.py](https://localhost:8080/#) in <module>()
     34 from google.protobuf.message import DecodeError
     35 
---> 36 from lucid.misc.io.reading import read_handle
     37 from lucid.misc.io.scoping import current_io_scopes, set_io_scopes
     38 from lucid.misc.io.saving import nullcontext

[/usr/local/lib/python3.7/dist-packages/lucid/misc/io/reading.py](https://localhost:8080/#) in <module>()
     32 from filelock import FileLock
     33 
---> 34 from lucid.misc.io.writing import write_handle
     35 from lucid.misc.io.scoping import scope_url
     36 

[/usr/local/lib/python3.7/dist-packages/lucid/misc/io/writing.py](https://localhost:8080/#) in <module>()
     26 from contextlib import contextmanager
     27 from urllib.parse import urlparse
---> 28 from tensorflow import gfile
     29 
     30 from lucid.misc.io.scoping import scope_url

ImportError: cannot import name 'gfile' from 'tensorflow' (/usr/local/lib/python3.7/dist-packages/tensorflow/__init__.py)

Also, the title of this issue should probably be updated to "Activation Atlas Google Colab notebook - ImportError: cannot import name 'gfile' from 'tensorflow'" or similar, if @themolco is still around. Thought about making a new issue for better visibility.

Fixed with https://github.com/tensorflow/lucid/issues/201#issuecomment-626165841