How can I get Features extracted by encoder

How i can create dataset of feature that my autoencoder learns or in other words where is the dataset reduce by encoder?

Would you like to save the features learned by the encoder?
Depending on the model, you could return the latent vectors with the final output and just concatenate them into a tensor, and save them using torch.save.

Yes, I am trying to save the features learned by the encoder. Thanks for the answer.