I was trying to figure out how Pytorch sorts the labels in the output tensor. I’ve understood that with Torchvision the class_to_idx creates the order given how elements are retrieved by the ImageFolder method and this will be the idx order in the output tensor.

On the other hand, as regard NLP I am very confused.

Is it done while defining the network here below in un-explicit way?

```
class CharRNN(nn.Module):
def __init__(self, tokens, n_hidden=256, n_layers=2,
drop_prob=0.5, lr=0.001):
super(CharRNN, self).__init__()
self.drop_prob = drop_prob
self.n_layers = n_layers
self.n_hidden = n_hidden
self.lr = lr
# creating character dictionaries
self.chars = tokens
self.int2char = dict(enumerate(self.chars))
self.char2int = {ch: ii for ii, ch in self.int2char.items()}
## TODO: define the LSTM
self.lstm = nn.LSTM(len(self.chars), n_hidden, n_layers,
dropout=drop_prob, batch_first=True)
self.dropout = nn.Dropout(drop_prob)
self.fc = nn.Linear(n_hidden, len(self.chars)+1000)
def forward(self, x, hidden):
''' Forward pass through the network.
These inputs are x, and the hidden/cell state `hidden`. '''
r_output, hidden = self.lstm(x, hidden)
out = self.dropout(r_output)
# Stack up LSTM outputs using view
# you may need to use contiguous to reshape the output
out = out.contiguous().view(-1, self.n_hidden)
out = self.fc(out)
# return the final output and the hidden state
return out, hidden
def init_hidden(self, batch_size):
''' Initializes hidden state '''
# Create two new tensors with sizes n_layers x batch_size x n_hidden,
# initialized to zero, for hidden state and cell state of LSTM
weight = next(self.parameters()).data
if (train_on_gpu):
hidden = (weight.new(self.n_layers, batch_size, self.n_hidden).zero_().cuda(),
weight.new(self.n_layers, batch_size, self.n_hidden).zero_().cuda())
else:
hidden = (weight.new(self.n_layers, batch_size, self.n_hidden).zero_(),
weight.new(self.n_layers, batch_size, self.n_hidden).zero_())
return hidden
```