PortaSpeech / mfa_usr /mfa_train_config.yaml
RayeRen's picture
init
d1b91e7
beam: 10
retry_beam: 40
features:
type: "mfcc"
use_energy: false
frame_shift: 10
training:
- monophone:
num_iterations: 40
max_gaussians: 1000
subset: 0
boost_silence: 1.25
- triphone:
num_iterations: 35
num_leaves: 2000
max_gaussians: 10000
cluster_threshold: -1
subset: 0
boost_silence: 1.25
power: 0.25
- lda:
num_leaves: 2500
max_gaussians: 15000
subset: 0
num_iterations: 35
features:
splice_left_context: 3
splice_right_context: 3
- sat:
num_leaves: 2500
max_gaussians: 15000
power: 0.2
silence_weight: 0.0
fmllr_update_type: "diag"
subset: 0
features:
lda: true
- sat:
num_leaves: 4200
max_gaussians: 40000
power: 0.2
silence_weight: 0.0
fmllr_update_type: "diag"
subset: 0
features:
lda: true
fmllr: true
# - monophone:
# num_iterations: 40
# max_gaussians: 1000
# boost_silence: 1.0
#
# - triphone:
# num_iterations: 35
# num_leaves: 3100
# max_gaussians: 50000
# cluster_threshold: 100
# boost_silence: 1.0
# power: 0.25
#
# - sat:
# num_leaves: 3100
# max_gaussians: 50000
# power: 0.2
# silence_weight: 0.0
# cluster_threshold: 100
# fmllr_update_type: "full"