How to handle batch size errors with custom loss functions?

Question

I cannot train a custom loss function when batch_size > 1 due to sizing errors.

class API_Network(object):
    def __init__(self):
        self.model = sel