module 'tensorflow.compat.v2' has no attribute '_internal'

[mahesaAtma]

Hi all, really quick question, i get these following error :

2021-06-02 03:36:48.320113: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
Using TensorFlow backend.
Traceback (most recent call last):
  File "keras-retinanet/keras_retinanet/bin/train.py", line 553, in <module>
    main()
  File "keras-retinanet/keras_retinanet/bin/train.py", line 469, in main
    backbone = models.backbone(args.backbone)
  File "keras-retinanet/keras_retinanet/bin/../../keras_retinanet/models/__init__.py", line 57, in backbone
    from .resnet import ResNetBackbone as b
  File "keras-retinanet/keras_retinanet/bin/../../keras_retinanet/models/resnet.py", line 18, in <module>
    import keras_resnet
  File "/usr/local/lib/python3.7/dist-packages/keras_resnet/__init__.py", line 1, in <module>
    from . import layers
  File "/usr/local/lib/python3.7/dist-packages/keras_resnet/layers/__init__.py", line 1, in <module>
    from ._batch_normalization import BatchNormalization
  File "/usr/local/lib/python3.7/dist-packages/keras_resnet/layers/_batch_normalization.py", line 1, in <module>
    import keras
  File "/usr/local/lib/python3.7/dist-packages/keras/__init__.py", line 12, in <module>
    from . import initializers
  File "/usr/local/lib/python3.7/dist-packages/keras/initializers/__init__.py", line 124, in <module>
    populate_deserializable_objects()
  File "/usr/local/lib/python3.7/dist-packages/keras/initializers/__init__.py", line 49, in populate_deserializable_objects
    LOCAL.GENERATED_WITH_V2 = tf.__internal__.tf2.enabled()
AttributeError: module 'tensorflow.compat.v2' has no attribute '__internal__'

Bassically all i do just installing all package required and execute cython, also i do Keras 2.4 and Tensorflow 2.3.0 as what the repository described. And after installing all packages, i tried to do the training as follow :

!keras-retinanet/keras_retinanet/bin/train.py --freeze-backbone --random-transform\
  --backbone resnet152 \
  --batch-size 8 \
  --steps 610 \
  --snapshot-path gdrive/MyDrive/Tugas_Akhir/snaphots \
  --tensorboard-dir gdrive/MyDrive/Tugas_Akhir/snaphots \
  csv gdrive/MyDrive/Tugas_Akhir/annotation.csv gdrive/MyDrive/Tugas_Akhir/classes.csv \
  --val-annotations gdrive/MyDrive/Tugas_Akhir/annotation_testing.csv

anyone know with this? i really need your help guys, please.

Thanks in advance.

[mahesaAtma]

For addition details, i ran it on google collab, about 2 weeks ago, i ran it, and it just working fine, but unfortunately, when i re-run this code,it givesme those errors

[syln21]

I guess it's because google colab upgrade tensorflow 2.5 , you should check issue #1537

[mahesaAtma]

I guess it's because google colab upgrade tensorflow 2.5 , you should check issue #1537

I did install follow versions : 2.1.0 2.2.0 2.3.0 2.3.1 2.3.2 2.4.0 and it gave me error log from the tensorflow itself. And yes, as in the requirements.txt, it will install newest tensorflow, which is currently 2.5.0

But unfortunately, i think keras retinanet is not supported yet for tensorflow 2.5.0? i did the training with 2.5.0, but it gives me 0 mAP with this error message

WARNING:tensorflow:
The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_116), but
are not present in its tracked objects:
  <tf.Variable 'bn4b26_branch2b/gamma:0' shape=(256,) dtype=float32>
  <tf.Variable 'bn4b26_branch2b/beta:0' shape=(256,) dtype=float32>
It is possible that this is intended behavior, but it is more likely
an omission. This is a strong indication that this layer should be
formulated as a subclassed Layer rather than a Lambda layer.

[kirthan0899]

Figured it out.
You just need to uninstall all keras and tensorflow packages and install the versions supported by Retinanet.

!pip uninstall keras -y
!pip uninstall keras-nightly -y
!pip uninstall keras-Preprocessing -y
!pip uninstall keras-vis -y
!pip uninstall tensorflow -y

!pip install tensorflow==2.3.0
!pip install keras==2.4

Just add this to the top of your colab notebook and restart your runtime once its completed and then you are good to go.

Hope this helps.

[mahesaAtma]

@kirthan0899 mannnn you just saved my life, tried dozens of options out there, reinstall, change python version, install anaconda, and other stuff. really want to know, how did you get those answer??

once again, thanksss! love you to the moon! haha

[kirthan0899]

tried dozens of options out there, reinstall, change python version, install anaconda, and other stuff.

Even I tried all those options and got fed up lol.
And later i went through some open issues and found this.

They had told to uninstall all packages and reinstall required packages, instead I just tried removing keras and tensorflow and it worked.

Anyway glad I could help you out :)

[S16201512]

Figured it out.
You just need to uninstall all keras and tensorflow packages and install the versions supported by Retinanet.

!pip uninstall keras -y
!pip uninstall keras-nightly -y
!pip uninstall keras-Preprocessing -y
!pip uninstall keras-vis -y
!pip uninstall tensorflow -y

!pip install tensorflow==2.3.0
!pip install keras==2.4

Just add this to the top of your colab notebook and restart your runtime once its completed and then you are good to go.

Hope this helps.

I love you，gay.
So good solution!
Thank you

[MEPOO13]

Figured it out.
You just need to uninstall all keras and tensorflow packages and install the versions supported by Retinanet.

!pip uninstall keras -y
!pip uninstall keras-nightly -y
!pip uninstall keras-Preprocessing -y
!pip uninstall keras-vis -y
!pip uninstall tensorflow -y

!pip install tensorflow==2.3.0
!pip install keras==2.4

Just add this to the top of your colab notebook and restart your runtime once its completed and then you are good to go.

Hope this helps.

Thanks Dear

[mantavya]

Figured it out.
You just need to uninstall all keras and tensorflow packages and install the versions supported by Retinanet.

!pip uninstall keras -y
!pip uninstall keras-nightly -y
!pip uninstall keras-Preprocessing -y
!pip uninstall keras-vis -y
!pip uninstall tensorflow -y

!pip install tensorflow==2.3.0
!pip install keras==2.4

Just add this to the top of your colab notebook and restart your runtime once its completed and then you are good to go.

Hope this helps.

it worked for me thanks!

[tulbureandreit]

Figured it out.
You just need to uninstall all keras and tensorflow packages and install the versions supported by Retinanet.

!pip uninstall keras -y
!pip uninstall keras-nightly -y
!pip uninstall keras-Preprocessing -y
!pip uninstall keras-vis -y
!pip uninstall tensorflow -y

!pip install tensorflow==2.3.0
!pip install keras==2.4

Just add this to the top of your colab notebook and restart your runtime once its completed and then you are good to go.

Hope this helps.

SAVED MY A**. thank you

[ZixianReid]

Works for me. Thanks!

[hoangnkust]

Figured it out. You just need to uninstall all keras and tensorflow packages and install the versions supported by Retinanet.

!pip uninstall keras -y !pip uninstall keras-nightly -y !pip uninstall keras-Preprocessing -y !pip uninstall keras-vis -y !pip uninstall tensorflow -y

!pip install tensorflow==2.3.0 !pip install keras==2.4

Just add this to the top of your colab notebook and restart your runtime once its completed and then you are good to go.

Hope this helps.

Do you know ???, you just saved my life !!!. Respect Sir !!!

[Abdodt]

importing it as tensorflow.keras.layers ....... will also work
just include " tensorflow. " when importing.

[Anvi98]

Best answer !

[amzker]

🤣

[NhatTamNL]

Figured it out. You just need to uninstall all keras and tensorflow packages and install the versions supported by Retinanet.

!pip uninstall keras -y !pip uninstall keras-nightly -y !pip uninstall keras-Preprocessing -y !pip uninstall keras-vis -y !pip uninstall tensorflow -y

!pip install tensorflow==2.3.0 !pip install keras==2.4

Just add this to the top of your colab notebook and restart your runtime once its completed and then you are good to go.

Hope this helps.

Thanks Dear

[hamedaligha]

I upgraded tensorflow-probability to version 0.24.0 then installed tf-keras package ---> problem is solved!

[RubTalha]

as @hamedaligha commented:

https://stackoverflow.com/a/78233592/13086128