Import error with tf-nightly and tf-keras on CPU

Keras
, ,
1

Installed tf-nightly and tf-keras in a new env
python version 3.11.16

import tf_keras
Traceback (most recent call last):
File “”, line 1, in
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/init.py”, line 3, in
from tf_keras import internal
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/internal/init.py”, line 6, in
from tf_keras.internal import models
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/internal/models/init.py”, line 3, in
from tf_keras.src.models.cloning import clone_and_build_model
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/init.py”, line 21, in
from tf_keras.src import applications
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/applications/init.py”, line 18, in
from tf_keras.src.applications.convnext import ConvNeXtBase
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/applications/convnext.py”, line 33, in
from tf_keras.src.engine import sequential
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/engine/sequential.py”, line 24, in
from tf_keras.src.engine import functional
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/engine/functional.py”, line 33, in
from tf_keras.src.engine import training as training_lib
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/engine/training.py”, line 48, in
from tf_keras.src.saving import saving_api
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/saving/saving_api.py”, line 25, in
from tf_keras.src.saving.legacy import save as legacy_sm_saving_lib
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/saving/legacy/save.py”, line 27, in
from tf_keras.src.saving.legacy.saved_model import load_context
File “/nfs/site/home/xyzd/tfv2_p311/lib/python3.11/site-packages/tf_keras/src/saving/legacy/saved_model/load_context.py”, line 68, in
tf.internal.register_load_context_function(in_load_context)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: module ‘tensorflow._api.v2.compat.v2.internal’ has no attribute ‘register_load_context_function’. Did you mean: ‘register_call_context_function’?

3

Hi @jojivk, Even though you installed tf-keras you can import keras using import keras. If you want to install keras-core you can install it using !pip install keras_core and import it using import keras_core as keras. Thank You.

5

Hi,

Thanks for reporting the issue.

Since the migration, there were many changes made, the error you are getting is due to one such change here Change only affects internal TF. · keras-team/tf-keras@cff6ac9 · GitHub.

To get rid of the error, you need to install tf-keras-nightly as well as tf-nightly.

Note that, tf-keras-nightly is legacy Keras code, to use the Keras 3 with multi-backend support, use keras-nightly and import Keras directly.