Error from loading a Keras model

[peterwu19881230]

Hi, dear developers:

When I was opening the following notebook in colab:
https://colab.research.google.com/github/ImagingDataCommons/IDC-Examples/blob/master/notebooks/lung_nodules_demo.ipynb

I wasn't able to load the model based on the json file: dp_arch.json

The error message is:
TypeError: init() missing 2 required positional arguments: 'filters' and 'kernel_size'

There are some instruction that relates to this. However, I am confused with the wording "change all the layers" at this time
"""
N.B. any warnings are due to the fact that the model was developed for
Keras 1, and the config file has been converted in a Keras-2-compatible file

Nonetheless, Keras 2 uses different naming conventions/def.s, so in order to
get rid of the warnings one should change all the layers def.s in the JSON file
"""

May I get some help from you?

Thanks!

[wlongabaugh]

@denbonte Can you provide guidance here?

[denbonte]

Dear Peter,

TypeError: init() missing 2 required positional arguments: 'filters' and 'kernel_size'

We know about this error and discussed about it (and how to fix it!) in this thread on the IDC forum. Unfortunately, Google removed native support for TensorFlow 1.x from Colab - so there is nothing we can do besides manually downgrade the TensorFlow version. This will affect ALL the models developed with TF 1.x running on Colab - not only the one we are using!

If you have any more questions, feel free to continue the thread here or join the discussion on the IDC forum!