I am trying to follow the third NLP tutorial here. I’m having trouble understanding the relationship between the input and hidden feature sizes in the encoder module.
The relevant code is:
class EncoderRNN(nn.Module):
def __init__(self, input_size, hidden_size):
super(EncoderRNN, self).__init__()
self.hidden_size = hidden_size
self.embedding = nn.Embedding(input_size, hidden_size)
self.gru = nn.GRU(hidden_size, hidden_size)
def forward(self, input, hidden):
embedded = self.embedding(input).view(1, 1, -1)
output = embedded
output, hidden = self.gru(output, hidden)
return output, hidden
def initHidden(self):
return torch.zeros(1, 1, self.hidden_size, device=device)
Won’t the shape of embedded
be (1, 1, input_size * hidden_size)
? If so, how can this be fed into a GRU
expecting and input of shape (1, 1, hidden_size)
?
I’m trying something like:
model = EncoderRNN(input_size=5, hidden_size=2)
model(input=torch.tensor([[[1,2,3,4,0]]]),
hidden=torch.tensor([[[0,0,0,0,0,0]]]))
but I get RuntimeError: input.size(-1) must be equal to input_size. Expected 2, got 10
, as expected.
So I am probably not calling model
correctly. What am I missing?