cmds2/run_FeatExt_Kaldi.py  --  Extract Features from a Trained Model and Save Features to the Kaldi Format
----------------------------------------------------------------------------------------------------------------------

Arguments

argument
meaning
default value
--in-scp-file             
path to the .scp file for network input features
required                    
--out-ark-filepath to the (Kaldi-formatted) .ark file to store the network activationrequired
--nnet-param path to model parameters required
--nnet-cfg
path to model config
required
--layer-index    
the index of the layer from which features are generated. Layers are 0-based indexed, starting from the first hidden layer. Set it to "-1" when you want to do classification. That is, we extract features from the final softmax layer.
required


Example

python cmds2/run_FeatExt_Kaldi.py --in-scp-file input.1.scp \
                                  --out-ark-file feats.1.ark \
                                  --nnet-param cnn.param --nnet-cfg cnn.cfg \
                                  --layer-index 1