Incompatibility in 'tf.GradientTape.watch' of TensorFlow 2.17 in Keras 3.4.1

[yajuna]

I read the issue 19155 (#19155), but still have problem

I am trying to perform gradient descent on the model.trainable variables, but have errors regarding model.trainable_variables

Tensorflow version is 2.17.0
keras version is 3.4.1

def get_grad(model, X_train, data_train):

with tf.GradientTape(persistent=True) as tape:
# This tape is for derivatives with
# respect to trainable variables
tape.watch(model.trainable_variables.value) ###added .value from issue 19155
loss = compute_loss(model, X_train, data_train)

g = tape.gradient(loss, model.trainable_variables.value) #
del tape

return loss, g

###################
Error:

AttributeError: in user code:

File "", line 7, in train_step *
loss, grad_theta = get_grad(model, X_train, data_train)
File "", line 6, in get_grad *
tape.watch(model.trainable_variables.value)

AttributeError: 'list' object has no attribute 'value'

[mehtamansi29]

Hi @yajuna -

Thanks for reporting this issue. Can you provide more details about your compute loss function and dummy datasets to reproduce this issue ?

[github-actions]

This issue is stale because it has been open for 14 days with no activity. It will be closed if no further activity occurs. Thank you.

[github-actions]

This issue was closed because it has been inactive for 28 days. Please reopen if you'd like to work on this further.

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No