Converting placeholder from tensorflow to pytorch

1

I am a newbie programmer in pytorch. I am trying to convert this code from Approximate Inference for Deep Latent Gaussian Mixtures paper written in tensorflow to pytorch. I have a difficult time finding out how I can translate tf.placeholder syntax from tensorflow to pytorch explicitly for this part of the aforementioned code:
line 84 of gaussMMVAE_collapsed.py script

class GaussMMVAE(object):
    def __init__(self, hyperParams):

        self.X = tf.placeholder("float", [None, hyperParams['input_d']])
        self.prior = hyperParams['prior']
        self.K = hyperParams['K']

        self.encoder_params = self.init_encoder(hyperParams)
        self.decoder_params = self.init_decoder(hyperParams)

I have read many posts that there isn’t any one to one function for placeholder but given the above example code, I reckon there should be a case by case solution. Thanks in advance for any instructive suggestion.

2

As you’ve already read, there are not placeholders in PyTorch and you can directly use tensors (similar to how you would use numpy). I don’t know why the tf.placeholder is assigned to self.X, but your translated code in PyTorch should just accept an input tensor in the forward(self, x) method and could use x directly in all operations.

1 Like