def validation_epoch_end(self, outputs):
if self.logger is not None and self.logger.experiment is not None:
logger = self.logger.experiment
if isinstance(self.logger, LoggerCollection):
for logger in self.logger:
if isinstance(logger, TensorBoardLogger):
logger = logger.experiment
break
if isinstance(logger, TensorBoardLogger):
tacotron2_log_to_tb_func(
logger,
outputs[0].values(),
self.global_step,
tag="val",
log_images=True,
add_audio=False,
)
elif isinstance(logger, WandbLogger):
tacotron2_log_to_wandb_func(
logger,
outputs[0].values(),
self.global_step,
tag="val",
log_images=True,
add_audio=False,
)
avg_loss = torch.stack([
x['val_loss'] for x in outputs
]).mean() # This reduces across batches, not workers!
self.log('val_loss', avg_loss)
def validation_epoch_end(self, outputs):
if self.logger is not None and self.logger.experiment is not None:
tacotron2_log_to_tb_func(
self.logger.experiment,
outputs[0].values(),
self.global_step,
tag="val",
log_images=True,
add_audio=False,
)
avg_loss = torch.stack([x['val_loss'] for x in outputs]).mean()
self.log('val_loss', avg_loss)