get_lda_nnet

aligner.multiprocessing.get_lda_nnet(config, align_directory, num_jobs)[source]

Multiprocessing function that extracts training examples and does LDA transformation

See:

for more details on the Kaldi binary this runs.

Also see https://github.com/kaldi-asr/kaldi/blob/master/egs/wsj/s5/steps/nnet2/get_lda.sh for the original bash script that this function was based on.

Parameters:
nnet_dir : str

Directory of nnet training

ali_dir : str

Directory of previous alignment

ivector_dir : str

Directory of the extracted i-vectors

training_dir : str

Directory of combined “split” files (e.g. utt2spk, cmvn.scp)

split_dir : str

Directory of training data split into the number of jobs

feats : str

Path to features

sil_phones : str

Colon-separated list of silence phones

config : NnetBasicConfig

Configuration object for training

num_jobs : int

The number of processes to use in calculation