Keras-yolo3: module 'keras.backend' has no attribute 'control_flow_ops'

[smalldroid]

I get this problem when run code on google colab. Please show me how to fix it? Thanks!

[datduonguva]

The code is old. It would not work with Keras 2.2.5. You can switch to
Keras 2.2.4 or lower, or change K.control_flow_ops.while_loop to
tf.while_loop. I havent tried this, though.

On Sat, Aug 31, 2019, 5:05 AM smalldroid [e-mail] wrote:

I get this problem when run code on google colab. Please show me how to
fix it? Thanks!

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/qqwweee/keras-yolo3/issues/544?email_source=notifications&email_token=AFSNTUOYUPH6ZRL2TC27GJDQHIX6PA5CNFSM4ISTCQM2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HISOZOA,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AFSNTUPXSQM6N5WV2R6UK5LQHIX6PANCNFSM4ISTCQMQ
.

[smalldroid]

Thanks for you answer! I tried to change to tf.while_loop but still not get luck.

[perlie-f]

tf.while_loop working here, using tensorflow 1.14.0

[JarvisPeng]

I also met this problem on google colab. Here's a solution:

add
from .load_backend import control_flow_ops
from .load_backend import set_image_dim_ordering
to /usr/local/lib/python3.6/dist-packages/keras/backend/__init__.py

code:
os.chdir('/usr/local/lib/python3.6/dist-packages/keras/backend')
!cat __init__.py
then copy and overwrite it:
%%writefile __init__.py

[ianchute]

Run the following Bash script:

echo "from .load_backend import control_flow_ops
from .load_backend import set_image_dim_ordering
$(cat /usr/local/lib/python3.6/dist-packages/keras/backend/__init__.py)" > /usr/local/lib/python3.6/dist-packages/keras/backend/__init__.py

[smalldroid]

Following Jarvis Peng to solve this problem. Thanks all!

[bubble0000]

I also get this problem on google colab.how to overwrite file on google colab.It seems that it doesn't really change the _init_.py .Because when I see this file again,it remains the same.And I try to change the version of tensorflow and keras,but I don't know why it remain prints the original version.When I using pip list,it show the version I try to install

[SunainaGoyal]

I even changed the __init.py. But the error still persist.
Any suggestion????

[ihmon]

I think that control_flow_ops belongs to tensorflow, thus I added following to __init__.py in backend folder in keras folder. Then it worked.

elif backend() == 'tensorflow':
from tensorflow.python.ops import control_flow_ops

Im not sure about set_image_dim_ordering, though.

[mammadjv]

I finally solved the problem by downgrading my Keras from version 2.3 to 2.1.5.

[potoo0]

I think that control_flow_ops belongs to tensorflow, thus I added following to init.py in backend folder in keras folder. Then it worked.

elif backend() == 'tensorflow':
from tensorflow.python.ops import control_flow_ops

Im not sure about set_image_dim_ordering, though.

thx, I ignored set_image_dim_ordering, it works on tf==1.15.2 keras==2.3.0

[keloy19]

i used this
%tensorflow_version 2.x
!pip uninstall -y tensorflow
!pip install tensorflow-gpu==1.14.0

and it worked