Save transfer learning weights to cache file

how do i save inception model transfer learning weights to cache file for reuse. i need use it for second LSTM model.

You could save the state_dict following the serialization semantics.
Would that work for you or what kind of cache would you like to use?

i need use final dense layer 2048 elements save as vector for fast uses as initial state for LSTM(h0).

I assume you are not running both models in the same script?
In that case, you could just store these features using and load them for your LSTM using torch.load.

Yes. of course, i am using it separately. Thank you