ModuleNotFoundError: No module named 'keras.legacy_tf_layers'

[akshatsingh22]

Is there an existing issue for this?

• I have searched the existing issues

Bug description

ModuleNotFoundError: No module named 'keras.legacy_tf_layers'

Google Colab!

Operating System

operating system: Google Colab

DeepLabCut version

dlc version: 2.3.5

DeepLabCut mode

single animal

Device type

gpu: Nvidia Tesla V100

Steps To Reproduce

from google.colab import drive
drive.mount('/content/drive')

!pip install deeplabcut

import os
import deeplabcut

!unzip 'MyProject.zip'

config_path = '/content/MyProject/config.yaml'

deeplabcut.create_training_dataset(
config_path,
num_shuffles=1,
net_type="resnet_50",
augmenter_type='imgaug'
)

deeplabcut.train_network(config_path,
shuffle=1,
trainingsetindex=0,
gputouse=None,
max_snapshots_to_keep=3,
autotune=False,
displayiters=10,
saveiters=1000,
maxiters=10000,
allow_growth=True)

Relevant log output

/usr/lib/python3.10/importlib/_bootstrap.py in _gcd_import(name, package, level)

/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load(name, import_)

/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)

/usr/lib/python3.10/importlib/_bootstrap.py in _call_with_frames_removed(f, *args, **kwds)

/usr/lib/python3.10/importlib/_bootstrap.py in _gcd_import(name, package, level)

/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load(name, import_)

/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)

ModuleNotFoundError: No module named 'keras.legacy_tf_layers'

Anything else?

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[seesaw1992]

I ran into the same problem with Colab.

[n-poulsen]

Which version of TensorFlow do you have installed (and how did you install DeepLabCut)?

I believe this is an error linked with an incompatible version of Tensorflow being installed. You should install tensorflow version <=2.12, so one easy fix would be running pip install --upgrade "tensorflow<=2.10" (on Google Colab/Windows/Linux machines).

For reference, on Apple computers with M1/M2 chips, you would need to run pip install --upgrade "tensorflow-macos<2.13.0".

[akshatsingh22]

Hi @n-poulsen ,

Many thanks for the clarification. Looks like the above solution worked. Outlining steps for others. The bug came in place as google has updated colab to tensorflow 2.13.0.

from google.colab import drive
drive.mount('/content/drive')

!pip install deeplabcut

!pip install --upgrade "tensorflow<=2.10"

Rest steps same as mentioned in the bug report. Training started after that for me so I would say this issue can be closed. Thanks again for the help.