The dataloader, train_dataloader, does not have many workers which may be a bottleneck. Consider increasing the value of the num_workers argument`

[coffeecodeconverter]

For anyone that gets this warning message during training:

rank_zero_warn(
/home/PiperTTS/.venv/lib/python3.10/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:224: 
PossibleUserWarning: 
The dataloader, train_dataloader, does not have many workers which may be a bottleneck. 
Consider increasing the value of the `num_workers` argument` (try 16 which is the number of cpus on this machine) 
in the `DataLoader` init to improve performance.

can be resolved by editing the lightning.py found here:

/home/PiperTTS/src/python/piper_train/vits/lightning.py

within this class:

class VitsModel(pl.LightningModule):
    def __init__(
        self,
        num_symbols: int,
        num_speakers: int,
        # audio
        resblock="2",
        resblock_kernel_sizes=(3, 5, 7),
        resblock_dilation_sizes=(
            (1, 2),
            (2, 6),
            (3, 12),
        ),
        upsample_rates=(8, 8, 4),
        upsample_initial_channel=256,
        upsample_kernel_sizes=(16, 16, 8),
        # mel
        filter_length: int = 1024,
        hop_length: int = 256,
        win_length: int = 1024,
        mel_channels: int = 80,
        sample_rate: int = 22050,
        sample_bytes: int = 2,
        channels: int = 1,
        mel_fmin: float = 0.0,
        mel_fmax: Optional[float] = None,
        # model
        inter_channels: int = 192,
        hidden_channels: int = 192,
        filter_channels: int = 768,
        n_heads: int = 2,
        n_layers: int = 6,
        kernel_size: int = 3,
        p_dropout: float = 0.1,
        n_layers_q: int = 3,
        use_spectral_norm: bool = False,
        gin_channels: int = 0,
        use_sdp: bool = True,
        segment_size: int = 8192,
        # training
        dataset: Optional[List[Union[str, Path]]] = None,
        learning_rate: float = 2e-4,
        betas: Tuple[float, float] = (0.8, 0.99),
        eps: float = 1e-9,
        batch_size: int = 1,
        lr_decay: float = 0.999875,
        init_lr_ratio: float = 1.0,
        warmup_epochs: int = 0,
        c_mel: int = 45,
        c_kl: float = 1.0,
        grad_clip: Optional[float] = None,
        num_workers: int = 1,
        seed: int = 1234,
        num_test_examples: int = 5,
        validation_split: float = 0.1,
        max_phoneme_ids: Optional[int] = None,
        **kwargs,
    ):

change this default

num_workers: int = 1

to anything above 2, otherwise it still gives you the warning
i have 16 cores available, but as an initial test, only increased it to 4

num_workers: int = 4

and you'll no longer see the warning during training.