nnsvs.train_util

Specific utilities for training

Helper for training

`setup`	Setup for training
`save_checkpoint`	Save a checkpoint.
`num_trainable_params`	Count the number of trainable parameters in the model.

`batch_by_size`	Yield mini-batches of indices bucketed by size.
`collate_fn_default`	Create batch
`collate_fn_random_segments`	Collate function with random segments
`get_data_loaders`	Get data loaders for training and validation.

Plot acoustic parameters of parametric SVS

`note_segments`	Compute note segments (start and end indices) from log-F0
`compute_pitch_regularization_weight`	Compute pitch regularization weight given note segments
`compute_batch_pitch_regularization_weight`	Batch version of computing pitch regularization weight
`compute_distortions`	Compute distortion measures between predicted and ground-truth acoustic features
`ensure_divisible_by`	Ensure that the number of frames is divisible by N.