Question about Temporally Fully Connected LSTM in PyTorch

I am currently working on a project involving LSTM (Long Short-Term Memory) networks in PyTorch and have a question about the connectivity of LSTM layers, specifically regarding “temporally fully connected” connections.

I would like to confirm if my understanding is correct: Does PyTorch’s nn.LSTM implementation indeed have these temporally fully connected connections, where each time step in a sequence is connected to each hidden unit in the LSTM layer?

If so, this would align with the standard behavior of LSTM architectures. I want to ensure that I have the correct understanding of how PyTorch handles these connections within LSTM layers.

Suppose I have input_size=10 and hidden_size =100 and time_stamps=10, than would input at each time stamp map to each hidden state in pytorch?

Please clarify and reference any sources(if possible)…