def get_optimizer(self) -> List:
"""Initiate and return the GAN optimizers based on the config parameters.
It returnes 2 optimizers in a list. First one is for the generator and the second one is for the discriminator.
Returns:
List: optimizers.
"""
gen_parameters = chain(
self.text_encoder.parameters(),
self.posterior_encoder.parameters(),
self.flow.parameters(),
self.duration_predictor.parameters(),
self.waveform_decoder.parameters(),
)
# add the speaker embedding layer
if hasattr(
self, "emb_g"
) and self.args.use_speaker_embedding and not self.args.use_d_vector_file:
gen_parameters = chain(gen_parameters, self.emb_g.parameters())
# add the language embedding layer
if hasattr(self, "emb_l") and self.args.use_language_embedding:
gen_parameters = chain(gen_parameters, self.emb_l.parameters())
optimizer0 = get_optimizer(self.config.optimizer,
self.config.optimizer_params,
self.config.lr_gen,
parameters=gen_parameters)
optimizer1 = get_optimizer(self.config.optimizer,
self.config.optimizer_params,
self.config.lr_disc, self.disc)
return [optimizer0, optimizer1]
def get_optimizer(self) -> List:
"""Initiate and return the GAN optimizers based on the config parameters.
It returnes 2 optimizers in a list. First one is for the generator and the second one is for the discriminator.
Returns:
List: optimizers.
"""
optimizer1 = get_optimizer(
self.config.optimizer, self.config.optimizer_params, self.config.lr_gen, self.model_g
)
optimizer2 = get_optimizer(
self.config.optimizer, self.config.optimizer_params, self.config.lr_disc, self.model_d
)
return [optimizer1, optimizer2]
def get_optimizer(self):
return get_optimizer(self.config.optimizer,
self.config.optimizer_params, self.config.lr,
self)