I have a model where one of the layers is an Axon embedding layer. I want to get that layer results instead of the model original output.
Since I’m building the model myself, what I did was have two models, one where I stopped in the embedding layer, and the other that has the full model.
Then, I used the full model to train the model state, and afterwards, I used the embedding-only layer to do the predict.
This seems to give me the embedding layer values as I wanted. But I’m not sure if that is actually the correct way to do that. Is it? Or is there a more “proper” way?
