An operation has `None` for gradient for triplet_hard_loss

[laurensnaturalis]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 16.04
• TensorFlow version and how it was installed (source or binary): 2.1.0 binary
• TensorFlow-Addons version and how it was installed (source or binary): 0.9.1 binary
• Python version: 3.6
• Is GPU used? (yes/no): yes

Describe the bug

I get An operation has None for gradient. when I use tfa.losses.triplet_hard_loss in non-eager mode. categorical_crossentropy works fine as loss

Code to reproduce the issue

   base_model = InceptionResNetV2(weights='imagenet', include_top=True)
    model = Model(inputs=base_model.input, outputs=predictions)
   loss=tfa.losses.triplet_hard_loss(y, z, margin=margin, soft=False)
    
    model.compile(optimizer=RMSprop(lr=0.0001, rho=0.9, epsilon=1e-08, decay=0.0), loss=loss, metrics=['accuracy'])

model.fit_generator(train_iterator, steps_per_epoch, epochs=num_epochs,
                                callbacks=[plot_loss_callback, reduce_learningrate_callback, checkpointer],  # , tensorboard_callback],
                                validation_data=test_iterator,
                                validation_steps=np.ceil(len(label_validation) / float(batch_size)),
                                workers=4) 

Other info / logs

  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/legacy/interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/engine/training.py", line 1732, in fit_generator
    initial_epoch=initial_epoch)
  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/engine/training_generator.py", line 42, in fit_generator
    model._make_train_function()
  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/engine/training.py", line 316, in _make_train_function
    loss=self.total_loss)
  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/legacy/interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py", line 75, in symbolic_fn_wrapper
    return func(*args, **kwargs)
  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/optimizers.py", line 259, in get_updates
    grads = self.get_gradients(loss, params)
  File "/home/laurens/anaconda2/envs/naturalis3c/lib/python3.6/site-packages/keras/optimizers.py", line 93, in get_gradients
    raise ValueError('An operation has `None` for gradient. '
ValueError: An operation has `None` for gradient. Please make sure that all of your ops have a gradient defined (i.e. are differentiable). Common ops without gradient: K.argmax, K.round, K.eval.

[WindQAQ]

Hi @laurensnaturalis, could you provide the minimal runnable code snippet so that we could diagnose where the problem is easily? Thank you!