Balanced batch generator returns inconsistent class number

Question

I am using imblearn.keras.balanced_batch_generator in my CNN classification task. But the generator produces inconsistent classes for my data (I have 12 classes in total but it produces 10/11/12 classes when batches are yielded).

This is causing an error when I train my model, indicating that my output layer node does not match with the class number (which matches this post: CNN model Categorical error: logits and labels must be broadcastable: logits_size=[32,10] labels_size=[32,13])

Code:

gen = BalancedBatchGenerator(data[["image"]], data[["class"]],
                                sampler=RandomOverSampler(),
                                random_state = 10,
                                batch_size=32)
for x,y in gen:
    print(len(y.value_counts())) # check number of class, dataset is in dataframe format
    # sample output 
    # 12
    # 11
    # 10
    # 12 

history = model.fit_generator(
    generator = train_generator, 
    validation_data = val_generator,
    epochs = 50,
    verbose = 1,
    callbacks = callbacks
)

Error:

InvalidArgumentError:  logits and labels must be broadcastable: logits_size=[32,11] labels_size=[32,12]
    [[node categorical_crossentropy/softmax_cross_entropy_with_logits (defined at <ipython-input-23-af8ee67b3eab>:6) ]] [Op:__inference_train_function_78283]