Extract Embedding Model after training

[manuelhuber]

So we have a (shared) embedding network that receives 1 input and outputs 1 embedding. We then build a triplet loss network around it that takes 3 inputs and outputs the loss which we train to minimize the loss. After we've trained our triplet loss network I'd like to use the embedding network without the triplet network. How would we access the "inner" embedding network? My goal is to be able to generate 1 embedding for 1 input.

[winnechan]

@manuelhuber i met the same problem, have you solved it?