Как мне создать модель вывода для этой модели кодера-декодера? ⇐ Python

[Anonymous]

Код: Выделить всё

encoder_inputs = tfl.Input(shape=(65 , ))
encoder_embedding = tfl.Embedding(4862, 256, mask_zero=True)(encoder_inputs)
encoder_outputs, state_h, state_c = tfl.LSTM(200, return_state=True)(encoder_embedding)
encoder_states = [state_h, state_c]

encoder_model = tf.keras.Model(encoder_inputs, encoder_states)

decoder_inputs = tfl.Input(shape=(56 - 1 , ))
decoder_embedding = tfl.Embedding(4708, 256, mask_zero=True)(decoder_inputs)
decoder_lstm = tfl.LSTM(200, return_state=True, return_sequences=True)
decoder_outputs, _, _ = decoder_lstm(decoder_embedding, initial_state=encoder_states)
decoder_dense = tfl.Dense(Y_sequence_vocab_size, activation="softmax")
decoder_outputs = decoder_dense(decoder_outputs)

model = tf.keras.Model([encoder_inputs, decoder_inputs], decoder_outputs)

Для этого я создал модель вывода: часть кодера работает нормально, но часть декодера выдает ошибку.
Модель вывода кодера
Модель вывода кодера
сильный>

Код: Выделить всё

encoder_inputs = model_new.input[0]
encoder_outputs, state_h, state_c = model_new.layers[4].output
encoder_states = [state_h, state_c]
encoder_model = tf.keras.Model(encoder_inputs, encoder_states)

Модель вывода декодера[/b]

Код: Выделить всё

decoder_inputs = model_new.input[1]  # input_2
decoder_state_input_h = keras.Input(shape=(55, 200))
decoder_state_input_c = keras.Input(shape=(200,))
decoder_states_inputs = [decoder_state_input_h, decoder_state_input_c]
decoder_lstm = model_new.layers[5]
decoder_outputs, state_h_dec, state_c_dec = decoder_lstm(
decoder_inputs, initial_state=decoder_states_inputs
)
decoder_states = [state_h_dec, state_c_dec]
decoder_dense = model_new.layers[6]
decoder_outputs = decoder_dense(decoder_outputs)
decoder_model = keras.Model(
[decoder_inputs] + decoder_states_inputs, [decoder_outputs] + decoder_states
)

Вот какую ошибку я получаю

Код: Выделить всё

ValueError                                Traceback (most recent call last)
Cell In[26], line 6
4 decoder_states_inputs = [decoder_state_input_h, decoder_state_input_c]
5 decoder_lstm = model_new.layers[5]
----> 6 decoder_outputs, state_h_dec, state_c_dec = decoder_lstm(
7     decoder_inputs, initial_state=decoder_states_inputs
8 )
9 decoder_states = [state_h_dec, state_c_dec]
10 decoder_dense = model_new.layers[6]

File c:\Users\SK Mahesh\AppData\Local\Programs\Python\Python38\lib\site-packages\keras\src\layers\rnn\base_rnn.py:615, in RNN.__call__(self, inputs, initial_state, constants, **kwargs)
613 # Perform the call with temporarily replaced input_spec
614 self.input_spec = full_input_spec
--> 615 output = super().__call__(full_input, **kwargs)
616 # Remove the additional_specs from input spec and keep the rest. It
617 # is important to keep since the input spec was populated by
618 # build(), and will be reused in the stateful=True.
619 self.input_spec = self.input_spec[: -len(additional_specs)]

File c:\Users\SK Mahesh\AppData\Local\Programs\Python\Python38\lib\site-packages\keras\src\utils\traceback_utils.py:70, in filter_traceback..error_handler(*args, **kwargs)
67     filtered_tb = _process_traceback_frames(e.__traceback__)
68     # To get the full stack trace, call:
69     # `tf.debugging.disable_traceback_filtering()`
---> 70     raise e.with_traceback(filtered_tb) from None
71 finally:
72     del filtered_tb

File c:\Users\SK Mahesh\AppData\Local\Programs\Python\Python38\lib\site-packages\keras\src\backend.py:4822, in rnn(step_function, inputs, initial_states, go_backwards, mask, constants, unroll, input_length, time_major, zero_output_for_mask, return_all_outputs)
4819 time_steps_t = tf.shape(flatted_inputs[0])[0]
4821 for input_ in flatted_inputs:
-> 4822     input_.shape.with_rank_at_least(3)
4824 if mask is not None:
4825     if mask.dtype != tf.bool:

ValueError: Exception encountered when calling layer "lstm_1" (type LSTM).

Shape (55, None) must have rank at least 3

Call arguments received by layer "lstm_1" (type LSTM):
• inputs=['tf.Tensor(shape=(None, 55), dtype=float32)', 'tf.Tensor(shape=(None, 55, 200), dtype=float32)', 'tf.Tensor(shape=(None, 200), dtype=float32)']
• mask=None
• training=None
• initial_state=None

Я не знаю, что именно вызывает эту ошибку, как ее исправить или точно закодировать модель вывода из основной модели.

Подробнее здесь: https://stackoverflow.com/questions/792 ... oder-model