AttributeError: 'types.SimpleNamespace' object has no attribute 'get_registered_object' from return keras.saving.get_registered_object(config["registered_name"]) in keras_nlp\src\utils\preset_utils.py

[JingsaiLu]

want to run the demo in the local by following the getting started guide and config the kaggle API credentials with this doc but get error when run the setup 5: `import keras import keras_nlp import os

os.environ["KERAS_BACKEND"] = "jax"

gemma_lm = keras_nlp.models.GemmaCausalLM.from_preset("gemma_2b_en")`

what is the reason for this ? and how to fix that ?

[SuryanarayanaY]

Hi @jsplyy ,

It seems to be version issue for me. Could you please confirm the keras_nlp and keras version being used ? Please install latest versions of both and let us know the outcome. Thanks!

[JingsaiLu]

Hi @SuryanarayanaY I am using, Name: keras-nlp Version: 0.9.1 https://pypi.org/project/keras-nlp/ Name: keras Version: 2.10.0

try to update keras with command: pip upgrade keras it will be updated to 3.1.1,

but when try to update keras_nlp with command: pip install --upgrade keras_nlp keras-2.10.0 will be installed again. keras_nlp seems to depends on and install 2.10.0 keras?

I checked the keras-nlp doc https://github.com/keras-team/keras-nlp/tree/master it documented that both keras2 and keras3 are supported.

Which version should I install ?

[SuryanarayanaY]

keras_nlp seems to depends on and install 2.10.0 keras?

Hi @jsplyy ,

No, it don't have any dependency on Keras 2.10.0 . Please ensure all the dependencies in requirements.txt installed. Keras_nlp needs Tensorflow also. May be you might have TF2.10v which might be reason for this behaviour.

[dongex]

I also faced the same error and I have installed Tensorflow

[gustheman]

is this issue still happening? can someone confirm?

[jvrapp]

@gustheman I'm still having this issue. When installing keras_nlp==0.12.1 it installs keras==2.10.0 and tf 2.10.1. Also when using keras_nlp==0.14.1 it doesn't let to import to jupyter due missing ops module. Somehow it works when installing libaries in the following order:

!pip install -U -q keras-nlp !pip install -q -U keras>=3

[mattdangerw]

KerasNLP no longer supports Keras 2, we are focusing on Keras 3. This means a few things...

• The latest versions of KerasNLP or Keras do not work with TensorFlow 2.14 or earlier.
• For TensorFlow 2.15. You need the weird install line @jvrapp showed. This is because Tensorflow 2.15 will pull in Keras 2 during installation.

  !pip install -U -q keras-nlp
  !pip install -q -U keras>=3

  • For TensorFlow 2.16 and greater. Everything should work normally. Tensorflow 2.16 will pull in Keras 3. Everything cooperates.

So TL;DR, make sure you base installation has tensorflow >2.16 and tensorflow-text >2.16 and things should be fine.

[mattdangerw]

In a colab environment, this should be quite simple. Colab now comes with TensorFlow 2.17. So

!pip install keras-nlp will do the right thing and pull in all necessary dependancies.

With a base environment where Tensorflow is older, you might want to do something like this to make sure your tf is upgraded...

!pip install -U keras-nlp tensorflow tensorflow-text

Hope that helps!

[Gopi-Uppari]

Hi @JingsaiLu,

Could you please confirm if this issue is resolved for you with the above comment ? Please feel free to close the issue if it is resolved ?

Thank you.