#reshape

A character-level LSTM with batch_first=True and n_hidden = 256 processes a batch of 10 sequences of 50 characters over a vocabulary of 83 characters:

r_output, hidden = self.lstm(x, hidden)          # r_output: (10, 50, 256)
out = r_output.contiguous().view(-1, self.n_hidden)
out = self.fc(out)                               # fc = nn.Linear(256, 83)

What is the shape of out?