AttributeError: module 'tensorflow' has no attribute 'gfile'

Question

I trained a simple mnist model with tensorflow 2.0 on Google Colab and saved it in the .json format. Click here to check out the Colab Notebook where I've written the code. Then on running the command

!simple_tensorflow_serving --model_base_path="/" --model_platform="tensorflow"

It is showing the error AttributeError: module 'tensorflow' has no attribute 'gfile'

simple_tensorflow_serving helps in easily deploying trained tensorflow model into production.

Versions I'm using:

(1) TensorFlow - 2.0

(2) simple_tensorflow_serving - 0.6.4

Thank you in advance :)

I'm having the same issue with `FullTokenizer(vocab_file, do_lower_case)` in TF 2 — bachr, Jan 09 '20 at 22:20

score 43 · Answer 1 · answered May 12 '20 at 09:18

43

In 2.0, tf.gfile.* is replaced by tf.io.gfile.*.

when I get error:

  File "/Users/MRJ/anaconda3/envs/python37-tf2.1/lib/python3.7/
site-packages/object_detection/utils/label_map_util.py",
line 137, in load_labelmap

with tf.gfile.GFile(path, 'r') as fid:
  AttributeError: module 'tensorflow' has no attribute 'gfile'

1.Find label_map_util.py line 137.

2.Replace tf.gfile.GFile to tf.io.gfile.GFile

It's worked for me.

tensorflow issue #31315

answered May 12 '20 at 09:18

JiangKui

1,207
1
11
17

Did you have any errors for other commands in tensorflow? I want to know if changing this will cause any other issues – Aaron John Sabu Jul 01 '20 at 11:47
1

No. "In 2.0, tf.gfile.* is replaced by tf.io.gfile.*." So tf.io.gfile.GFile alway is right. – JiangKui Jul 22 '20 at 10:05

score 25 · Accepted Answer · answered Apr 09 '19 at 14:32

25

Simple Tensorflow Serving is not ready for Tensorflow 2.0, since it is using the old API. In Tensorflow 2.0 the gfile package has been moved into tf.io.

Then, you have to downgrade your Tensorflow instance to TF 1.13 use Simple Tensorflow Serving

answered Apr 09 '19 at 14:32

nessuno

26,493
5
83
74

How do you downgrade it? – 0xTomato Apr 30 '20 at 01:54
3

pip install tensorflow==1.13 – nessuno Apr 30 '20 at 05:40
1

Incase anyone gets here and is getting "ERROR: Could not find a version that satisfies the requirement tensorflow==1.13". Please use ----> pip install tensorflow==1.13.1 – inyang kpongette Apr 17 '21 at 23:39

score 4 · Answer 3 · edited Dec 05 '21 at 16:24

4

Instead, try

tf.io.gfile.GFile(
    name, mode='r'
)

i.e ".io" will solve your all problems instead of downgrading your tf

edited Dec 05 '21 at 16:24

Suraj Rao

29,388
11
94
103

answered Dec 05 '21 at 16:23

Animesh Nayak

97
1
2

score 0 · Answer 4 · edited Sep 20 '22 at 18:51

0

change tf.gfile.GFile to tf.io.gfile.GFile.

it works for this instance but not for all other files

example tf.io.gfile.FastGFile throws an error saying

AttributeError: module 'tensorflow._api.v2.io.gfile' has no attribute 'FastGFile'

edited Sep 20 '22 at 18:51

desertnaut

57,590
26
140
166

answered Nov 20 '21 at 23:45

Anachaba

76
4

Your answer could be improved with additional supporting information. Please [edit] to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers [in the help center](/help/how-to-answer). – no ai please Nov 21 '21 at 00:57

score 0 · Answer 5 · answered Aug 16 '22 at 05:17

This is deprecated in TensorFlow 2.0. You can use older tensorflow version to use this api like 1.13.1

In my older project I needed this api Created a virtual environment in ananconda with this version like:

conda create -n tf python=3.7 tensorflow=1.13.1 
// here more modules with specific version can be added

conda activate tf // to activate environment tf

(base) D:\ff\testM>  --> (tf) D:\ff\testM> 
// After this environment changed to (base) --> (tf)

AttributeError: module 'tensorflow' has no attribute 'gfile'

5 Answers5

Linked