An RNN is built to tag sentences of up to tokens.
| quantity | size |
|---|---|
| token embedding arriving at each step | values |
| hidden state | values |
| longest sentence the model handles | tokens |
Inside the cell, each of the hidden units receives every value of the arriving token vector and every value of the previous hidden state, and carries a bias of its own. Nothing else in the cell holds a learned value.
How many learned parameters does this cell contain in total?
Select all that apply.