CODEWITHSUNDEEP
pythonpytorch
theQman
theQman

Reputation: 1780

Expected dimension sizes for pytorch models

I'm struggling with understanding what sort of dimensions my pytorch model needs as input. My model setup is:

import torch
from torch import nn, tensor

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()
        self.rnn_b = nn.RNN(input_size=input_size, 
                           hidden_size=hidden_size,
                           num_layers=1,
                           bias=False)
        self.hidden_size = hidden_size
        self.linearout = nn.Linear(hidden_size, output_size)

    def forward(self, input, hidden):
        out, hidden = self.rnn_b(input, hidden)
        output = self.linearout(out)
        return output, hidden

    def initHidden(self):
        return torch.zeros(1, self.hidden_size)

model = RNN(1, 2, 1)

If I call model(tensor(25.), tensor([[0., 0.]])) then I get the error:

IndexError: dimension specified as 1 but tensor has no dimensions

If I call model(tensor([25.]), tensor([[0., 0.]])) then I get the error

IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)

Can someone please explain what's going on here? What is the correct way to format the data for the input to my model?

Upvotes: 2

Views: 553

Answers (1)

Horace
Horace

Reputation: 1054

You need to encapsulate more. This is because pytorch automatically allows batches:

model(tensor([[[25.]]]), tensor([[[0., 0.]]]))

Output:

(tensor([[[-0.7704]]], grad_fn=<AddBackward0>),
 tensor([[[1., 1.]]], grad_fn=<StackBackward>))

You can use multiple input, as a batch:

model(tensor([[[25.]], [[25.]]]), tensor([[[0., 0.]]]))

Output:

(tensor([[[0.0341]],

         [[0.0341]]], grad_fn=<AddBackward0>),
 tensor([[[ 0.9999, -1.0000]]], grad_fn=<StackBackward>))

Upvotes: 1

Related Questions