TypeError: init() missing 1 required positional argument: 'units' when using the NoisyDense Class

Question

I am trying to implement Noisy Nets in my model. I found a code on GitHub which is an implementation of NoisyDense Class. I used this class inside my model. Here the code: -


class Agent:

    def __init__(self, state_size, strategy="t-dqn", reset_every=1000, pretrained=False, model_name=None):
        self.strategy = strategy

        # agent config
        self.state_size = state_size        
        self.action_size = 3                
        self.model_name = model_name
        self.inventory = []
        self.buffer = []
        self.first_iter = True
        self.nstep = 5
        self.n_step_buffer = deque(maxlen = self.nstep)
        self.cnt = count()
        self.alpha = 0.6
        self.model_name = model_name
        self.gamma = 0.95 # affinity for long term reward
        self.epsilon = 1.0
        self.epsilon_min = 0.01
        self.epsilon_decay = 0.995
        self.learning_rate = 0.001
        self.loss = huber_loss
        self.custom_objects = {"huber_loss": huber_loss}  # important for loading the model from memory
        self.optimizer = Adam(lr=self.learning_rate)

        if pretrained and self.model_name is not None:
            self.model = self.load()
        else:
            self.model = self._model()

        # strategy config
        if self.strategy in ["t-dqn", "double-dqn"]:
            self.n_iter = 1
            self.reset_every = reset_every

            # target network
            self.target_model = clone_model(self.model)
            self.target_model.set_weights(self.model.get_weights())
    def _model(self):
         model = Sequential()
         model.add(Dense(units=128, activation="relu", input_dim=self.state_size))
         model.add(Dense(units=256, activation="relu"))
         model.add(Dense(units=256, activation="relu"))
         model.add(NoisyDense(128, activation="relu"))
         model.add(NoisyDense(self.action_size + 1, activation='linear'))
         model.compile(loss=self.loss, optimizer=self.optimizer)
         return model

But ran into an error, I don't seem to know how to solve it, also Check NoisyDense implementation. I looked for similar issues on StackOverflow and I didn't get any solution that worked for my code. Also, the same model works when I don't call the NoisyDense class but use the Dense layer instead. Train.py file in the traceback calls the agent class nothing more important I guess. The model used is a Keras model and its version is 2.3.1

Traceback (most recent call last):
  File "train.py", line 85, in <module>
    pretrained=pretrained, debug=debug)
  File "train.py", line 52, in main
    agent = Agent(window_size, strategy=strategy, pretrained=pretrained, model_name=model_name)
  File "C:\Users\Desktop\agent.py", line 158, in __init__
    self.target_model = clone_model(self.model)
  File "D:\anaconda\lib\site-packages\keras\models.py", line 255, in clone_model
    return _clone_sequential_model(model, input_tensors=input_tensors)
  File "D:\anaconda\lib\site-packages\keras\models.py", line 207, in _clone_sequential_model
    layers = [clone(layer) for layer in model.layers]
  File "D:\anaconda\lib\site-packages\keras\models.py", line 207, in <listcomp>
    layers = [clone(layer) for layer in model.layers]
  File "D:\anaconda\lib\site-packages\keras\models.py", line 205, in clone
    return layer.__class__.from_config(layer.get_config())
  File "D:\anaconda\lib\site-packages\keras\engine\base_layer.py", line 1179, in from_config
    return cls(**config)
TypeError: __init__() missing 1 required positional argument: 'units'

score 0 · Answer 1 · answered Mar 03 '20 at 06:55

0

Class NoisyDense on Github as you can see takes unit as argument. As you are not giving this argument. That's why the error is showing

answered Mar 03 '20 at 06:55

NaSir HuSSaiN

192
4
12

is the first parameter not considered as units by defaults ```model.add(NoisyDense(128, activation="relu")) model.add(NoisyDense(self.action_size + 1, activation='linear'))``` – Raj Shah Mar 04 '20 at 06:34
well I also tried specifying it as units still the same error ``` model.add(NoisyDense(units = 128, activation="relu", sigma_init = 0.02))``` – Raj Shah Mar 04 '20 at 06:40

score 0 · Answer 2 · answered Mar 04 '20 at 11:11

0

Thank you for your time but I got the answer from a similar question on StackOverflow

"ValueError: Unknown layer: ... " when calling copy.deepcopy(network) using Tensorflow

answered Mar 04 '20 at 11:11

Raj Shah

13
1
4

TypeError: __init__() missing 1 required positional argument: 'units' when using the NoisyDense Class

2 Answers2

TypeError: init() missing 1 required positional argument: 'units' when using the NoisyDense Class