mirror of
https://github.com/wassname/TTS.git
synced 2026-09-10 11:50:20 +08:00
add speaker encoder parameters in config.json, for more easy changes
This commit is contained in:
@@ -135,10 +135,10 @@ def main(args): # pylint: disable=redefined-outer-name
|
||||
global meta_data_eval
|
||||
|
||||
ap = AudioProcessor(**c.audio)
|
||||
model = SpeakerEncoder(input_dim=40,
|
||||
proj_dim=128,
|
||||
lstm_dim=384,
|
||||
num_lstm_layers=3)
|
||||
model = SpeakerEncoder(input_dim=c.model['input_dim'],
|
||||
proj_dim=c.model['proj_dim'],
|
||||
lstm_dim=c.model['lstm_dim'],
|
||||
num_lstm_layers=c.model['num_lstm_layers'])
|
||||
optimizer = RAdam(model.parameters(), lr=c.lr)
|
||||
|
||||
if c.loss == "ge2e":
|
||||
|
||||
Reference in New Issue
Block a user