Difference between using 1 LSTM with 2 features and 2 LSTMs with 1 feature

If there’s a sequential dataset with with 2 features (eg. a Pytorch tensor of shape (N,2) ) and I want to use an LSTM to generate more elements in the sequence what would be the difference between using a single LSTM torch.nn.LSTM(input_size=2) for both features, or 2 LSTMs torch.nn.LSTM(input_size=1) one for each feature?