RNNBlock + DenseBlock

[holinov]

I cant figure out hot to use RNNBlock + DenseBlock when RNNBlock has return_sequences=True
I tried:

    input_node = ak.Input()
    rnn = ak.RNNBlock(return_sequences=True)(input_node)
    dense = ak.DenseBlock()(rnn)
    class_node = ak.ClassificationHead()(dense)

And getting

ValueError: The last dimension of the inputs to `Dense` should be defined. Found `None`.

Adding ak.TemporalReduction does not help

    input_node = ak.Input()
    rnn = ak.RNNBlock(return_sequences=True)(input_node)
    flat = ak.TemporalReduction()(rnn)
    dense = ak.DenseBlock()(flat)
    class_node = ak.ClassificationHead()(dense)

Getting same error. Also I tried ak.Flatten

[haifeng-jin]

This is probably a bug. I don't see an error here.
Which line does it report the error?

[holinov]

It repeats this error 5 times

Traceback (most recent call last):
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\kerastuner\engine\hypermodel.py", line 104, in build
    model = self.hypermodel.build(hp)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\kerastuner\engine\hypermodel.py", line 64, in _build_wrapper
    return self._build(hp, *args, **kwargs)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\autokeras\graph.py", line 250, in build
    outputs = block.build(hp, inputs=temp_inputs)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\autokeras\engine\block.py", line 38, in _build_wrapper
    return super()._build_wrapper(hp, *args, **kwargs)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\kerastuner\engine\hypermodel.py", line 64, in _build_wrapper
    return self._build(hp, *args, **kwargs)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\autokeras\blocks\basic.py", line 133, in build
    output_node = layers.Dense(units)(output_node)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\tensorflow\python\keras\engine\base_layer.py", line 951, in __call__
    return self._functional_construction_call(inputs, args, kwargs,
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\tensorflow\python\keras\engine\base_layer.py", line 1090, in _functional_construction_call
    outputs = self._keras_tensor_symbolic_call(
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\tensorflow\python\keras\engine\base_layer.py", line 822, in _keras_tensor_symbolic_call
    return self._infer_output_signature(inputs, args, kwargs, input_masks)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\tensorflow\python\keras\engine\base_layer.py", line 862, in _infer_output_signature
    self._maybe_build(inputs)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\tensorflow\python\keras\engine\base_layer.py", line 2710, in _maybe_build
    self.build(input_shapes)  # pylint:disable=not-callable
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\tensorflow\python\keras\layers\core.py", line 1182, in build
    raise ValueError('The last dimension of the inputs to `Dense` '
ValueError: The last dimension of the inputs to `Dense` should be defined. Found `None`.
Invalid model 0/5

And then says

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File ".\src\models\tvoyfrut\autokeras\lstm_bug_min.py", line 172, in <module>
    run_train(config)
  File ".\src\models\tvoyfrut\autokeras\lstm_bug_min.py", line 122, in run_train
    model.fit(train_x, verbose=2, callbacks=[early_stopping], epochs=2000)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\autokeras\auto_model.py", line 273, in fit
    self.tuner.search(
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\autokeras\engine\tuner.py", line 178, in search
    self.hypermodel.build(hp)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\kerastuner\engine\hypermodel.py", line 64, in _build_wrapper
    return self._build(hp, *args, **kwargs)
  File "C:\Users\tvoyf\anaconda3\envs\libra\lib\site-packages\kerastuner\engine\hypermodel.py", line 112, in build
    raise RuntimeError(
RuntimeError: Too many failed attempts to build model.

[holinov]

Maybe you could tell me some workaround?

[haifeng-jin]

Maybe removing the DenseBlock? In the ClassificationHead, it also contains some dense layers.

[holinov]

getting same error

[muaz-asr]

Hi, did you find any solution to it? I am having a similar problem.