Hidden Size Error

[Brandon-EME]

Hi,

Apologies if this question seems rudimentary however when running train.py I am getting a hidden state size error for the RNN module within torch.nn. Reviewing the code it seems that the hidden state size defined in forward is based on the previous state, rather than the output required for the current state, is this correct. I'm quite surprised this issue isn't present for anyone else as it occurs just by my running train.py so don't know if I am misunderstanding something or if there is a version error somewhere.

Error in Ipython terminal:

File ~\AppData\Local\anaconda3\envs\ntmselfbuild\Lib\site-packages\torch\nn\modules\rnn.py:260 in check_hidden_size
raise RuntimeError(msg.format(expected_hidden_size, list(hx.size())))

RuntimeError: Expected hidden[0] size (1, 1, 100), got [1, 1, 29]

Many thanks.