Pretrained acoustic model as Feature extractor

[lale10]

Hi ! I would like to extract and save the features of the last layer of the pre-trained acoustic model (the one before the output layer). Any hints ?

Thanks