TriphoneConfig

class aligner.config.TriphoneConfig(**kwargs)[source]

Configuration class for triphone training

Scale options defaults to:

['--transition-scale=1.0', '--acoustic-scale=0.1', '--self-loop-scale=0.1']

If align_often is True in the keyword arguments, realign_iters will be:

[1, 5, 10, 15, 20, 25, 30, 35, 38]

Otherwise, realign_iters will be:

[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 14, 16, 18, 20, 23, 26, 29, 32, 35, 38]
Attributes:
num_itersint

Number of training iterations to perform, defaults to 35

scale_optslist

Options for specifying scaling in alignment

beamint

Default beam width for alignment, defaults = 10

retry_beamint

Beam width to fall back on if no alignment is produced, defaults to 40

max_iter_incint

Last iter to increase #Gauss on, defaults to 30

totgaussint

Total number of gaussians, defaults to 1000

boost_silencefloat

Factor by which to boost silence likelihoods in alignment, defaults to 1.0

realign_iterslist

List of iterations to perform alignment

stageint

Not used

powerfloat

Exponent for number of gaussians according to occurrence counts, defaults to 0.25

do_fmllrbool

Specifies whether to do speaker adaptation, defaults to False

num_statesint

Number of states in the decision tree, defaults to 3100

num_gaussint

Number of gaussians in the decision tree, defaults to 50000

cluster_thresholdint

For build-tree control final bottom-up clustering of leaves, defaults to 100

Attributes

inc_gauss_count

max_iter_inc

Methods