I get "TypeError: 'NoneType' object is not callable" when train model

[KhizarAziz]

Hi, Thankyou for the code,
When i run the code, everything works totally fine except this,

When i execute this line,
models.fit(train_gen, steps_per_epoch=len(trainset) // batch_size, epochs=250, callbacks=callbacks, validation_data=validation_gen, validation_steps=len(testset) // batch_size * 3)

i get this error:

TypeError: 'NoneType' object is not callable`

Kindly help with this, or please guide to a direction where i can debug the code and solve the issue, i shall be highly thankful.

[StevenBanama]

@KhizarAziz check your tensorflow version with my requirements in the repos which may save your time

[KhizarAziz]

The problem was related to the age_data_generator.
out2 = [candis.age.to_numpy(), np.array([two_point(x, category, interval) for x in candis.age.to_numpy()])] yield [imgs[:,0], imgs[:,1], imgs[:,2]], out2

I changed above code to:
out2 = np.array([two_point(x, category, interval) for x in candis.age.to_numpy()]) yield [imgs[:,0], imgs[:,1], imgs[:,2]], out2

This fixed problem, but more questions arose here.....

1. Why it fixed problem
2. why were you return [age_numpy_array, two_point_representation_numpy]

i am confused, kindly help!!

[StevenBanama]

The problem was related to the age_data_generator.
out2 = [candis.age.to_numpy(), np.array([two_point(x, category, interval) for x in candis.age.to_numpy()])] yield [imgs[:,0], imgs[:,1], imgs[:,2]], out2

I changed above code to:
out2 = np.array([two_point(x, category, interval) for x in candis.age.to_numpy()]) yield [imgs[:,0], imgs[:,1], imgs[:,2]], out2

This fixed problem, but more questions arose here.....

1. Why it fixed problem
2. why were you return [age_numpy_array, two_point_representation_numpy]

i am confused, kindly help!!

@KhizarAziz age prediction in origin paper need two stage , meanwhile the age loss was composed by bucket loss and age MAE.The first stage gets age feature divided by age bucket and then compact these features to gen age prediction. "candis.age.to_numpy()" was add to calculate the loss of MAE.