C5w4 The final transformer code exercise 8

Nafi · August 21, 2021, 3:17pm

In the transformer function where everything is put together, we invoke the call function for both the encoder and decoder block using encoder(…) and decoder(…). Can anyone tell me the logic behind why this works and why we don’t instead invoke the encoder.call() function? Is call supposed to be a built in class method ? from my understanding built in methods follow the (underscore)(underscore)method(underscore)(underscore) format

jonaslalin · August 21, 2021, 3:24pm

Remember we use subclassing. Also, __call__ invokes call:

github.com

keras-team/keras/blob/b0d33577387bb0397492d0ee0af5b0d15d7c8385/keras/engine/base_layer.py#L969

    
      
              if any(m is not None for m in tf.nest.flatten(mask)):
                raise TypeError('Layer ' + self.name + ' does not support masking, '
                                'but was passed an input_mask: ' + str(mask))
              # masking not explicitly supported: return None as mask.
              return None
            # if masking is explicitly supported, by default
            # carry over the input mask
            return mask
          
          
@traceback_utils.filter_traceback
          def __call__(self, *args, **kwargs):
            """Wraps `call`, applying pre- and post-processing steps.
          
          
  Args:
              *args: Positional arguments to be passed to `self.call`.
              **kwargs: Keyword arguments to be passed to `self.call`.
          
          
  Returns:
              Output tensor(s).
          
          
  Note:

Topic		Replies	Views
Python question calling the week 1 NLP with Attention Models general	3	40	June 25, 2024
W4 Assignment 1 Exercise 8: Question about the custom Transformer model's call method Sequence Models week-4	3	306	January 9, 2024
W4_Ex-4 \| (Encoder) - Stuck with no clue what to do next! Please help! Sequence Models	14	1670	November 9, 2022
Issues with computing self_mha_ouput Sequence Models	5	645	February 8, 2024
Week 4 assignment 1 UNQ_C4 The first argument to `Layer.call` must always be passed Sequence Models	1	629	April 1, 2022

C5w4 The final transformer code exercise 8

Related topics