Hi, I was trying to give python nested list as nn.LSTM’s input. I have converted the list to LongTensor. My code is as follows. May I ask if it is the case that nn.LSTM can only take FloatTensor as input but never LongTensor? Thank you.
import torch
import torch.nn as nn
from torch.autorgrad import Variable
size_vocab = 5
hidden_size = 5
seq_len = 10
lstm = nn.LSTM(size_vocab, hidden_size)
inputs = Variable(torch.randn(seq_len,1,size_vocab)
out, hidden = lstm(inputs, None) # this works
one_hot = [[0] * size_vocab] * seq_len
inputs2 = Variable(torch.LongTensor([one_hot]).transpose(0,1)
out2, hidden2 = lstm(inputs2, None) # this won't work