ValueError: Tried to convert 'y' to a tensor and failed. Error: None value not supported

[pairwiserr]

tf.__version__
'2.0.0'
tf.keras.__version__
'2.2.4-tf'

I'm running a very vanilla Sequential model API

model.fit(x_train_pad, y_train, verbose=2.0, validation_split=0.05, epochs=3, batch_size=64, shuffle=True)

where x_train_pad is a python list of tokens for some texts with pre-padding (0's in front of list). and y_train is a python list of 1.0 or 0.0 depending on whether the corresponding text in x_train_pad is positive or negative sentiment.

The ValueError is from grads = K.gradients(loss,params) in the get_gradients method of keras optimizer.py. Seems like tensorflow assigns a None value for gradient of value 0. But this is causing downstream error when I try fitting my Sequential model to my training dataset.