AttributeError: cannot assign module before Module.init() call

Question

I am getting the following error.

Traceback (most recent call last):
  File "main.py", line 63, in <module>
    question_classifier = QuestionClassifier(corpus.dictionary, embeddings_index, corpus.max_sent_length, args)
  File "/net/if5/wua4nw/wasi/academic/research_with_prof_chang/projects/question_answering/duplicate_question_detection/source/question_classifier.py", line 26, in __init__
    self.embedding = EmbeddingLayer(len(dictionary), args.emsize, args.dropout)
  File "/if5/wua4nw/anaconda3/lib/python3.5/site-packages/torch/nn/modules/module.py", line 255, in __setattr__
    "cannot assign module before Module.__init__() call")
AttributeError: cannot assign module before Module.__init__() call

I have a class as follows.

class QuestionClassifier(nn.Module):

     def __init__(self, dictionary, embeddings_index, max_seq_length, args):
         self.embedding = EmbeddingLayer(len(dictionary), args.emsize, args.dropout)
         self.encoder = EncoderRNN(args.emsize, args.nhid, args.model, args.bidirection, args.nlayers, args.dropout)
         self.drop = nn.Dropout(args.dropout)

So, when I run the following line:

question_classifier = QuestionClassifier(corpus.dictionary, embeddings_index, corpus.max_sent_length, args)

I get the above mentioned error. Here, EmbeddingLayer and EncoderRNN is a class written by me which inherits nn.module like the QuestionClassifier class.

What I am doing wrong here?

Asked By: Wasi Ahmad

||

Source

Answer 1

Looking at the pytorch source code for Module, we see in the docstring an example of deriving from Module includes:

 class Model(nn.Module):
        def __init__(self):
            super(Model, self).__init__()
            self.conv1 = nn.Conv2d(1, 20, 5)
            self.conv2 = nn.Conv2d(20, 20, 5)

So you probably want to call Module‘s init the same way in your derived class:

super(QuestionClassifier, self).__init__()

Answered By: juanpa.arrivillaga

Answer 2

Pytorch keeps track of the submodules(conv1, conv2) you will write in your custom Module. Under the hood, the graph corresponding to your Model is automatically built.

The nested Modules will be added to an OrderedDict _modules (initialized in nn.Module.__init__) See source(L69)

If nn.Module.__init__ is not called (self._modules would equal to None), when trying to add a Module, it will raise an error (no key can be added to None). See
source(L540-544)

Inspired from the doc:

 class CustomModule(nn.Module):
        def __init__(self):
            super(CustomModule, self).__init__() # Initialize self._modules as OrderedDict
            self.conv1 = nn.Conv2d(1, 20, 5)     # Add key conv1 to self._modules
            self.conv2 = nn.Conv2d(20, 20, 5)    # Add key conv2 to self._modules

Answered By: x0s

Answer 3

This usually happens when super class’s init has not been called. In this case one should start their Neural network class with super.__init__() call. The code would look like this:

class QuestionClassifier(nn.Module):
    def __init__(self, dictionary, embeddings_index, max_seq_length, args):
       super().__init__()

This super’s init call should be within this class’s init code.

Answered By: Anurag

AttributeError: cannot assign module before Module.init() call

Question:

Answers: