train_ms_jp_extra.py

import argparse
import datetime
import gc
import os
import platform

import torch
import torch.distributed as dist
from torch.cuda.amp import GradScaler, autocast
from torch.nn import functional as F
from torch.nn.parallel import DistributedDataParallel as DDP
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from tqdm import tqdm

# logging.getLogger("numba").setLevel(logging.WARNING)
import commons
import default_style
import utils
from common.log import logger
from common.stdout_wrapper import SAFE_STDOUT
from config import config
from data_utils import (
    DistributedBucketSampler,
    TextAudioSpeakerCollate,
    TextAudioSpeakerLoader,
)
from losses import WavLMLoss, discriminator_loss, feature_loss, generator_loss, kl_loss
from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
from models_jp_extra import (
    DurationDiscriminator,
    MultiPeriodDiscriminator,
    SynthesizerTrn,
    WavLMDiscriminator,
)
from text.symbols import symbols

torch.backends.cuda.matmul.allow_tf32 = True
torch.backends.cudnn.allow_tf32 = (
    True  # If encontered training problem,please try to disable TF32.
)
torch.set_num_threads(1)
torch.set_float32_matmul_precision("medium")
torch.backends.cuda.sdp_kernel("flash")
torch.backends.cuda.enable_flash_sdp(True)
torch.backends.cuda.enable_mem_efficient_sdp(
    True
)  # Not available if torch version is lower than 2.0
global_step = 0


def run():
    # Command line configuration is not recommended unless necessary, use config.yml
    parser = argparse.ArgumentParser()
    parser.add_argument(
        "-c",
        "--config",
        type=str,
        default=config.train_ms_config.config_path,
        help="JSON file for configuration",
    )
    parser.add_argument(
        "-m",
        "--model",
        type=str,
        help="数据集文件夹路径，请注意，数据不再默认放在/logs文件夹下。如果需要用命令行配置，请声明相对于根目录的路径",
        default=config.dataset_path,
    )
    parser.add_argument(
        "--assets_root",
        type=str,
        help="Root directory of model assets needed for inference.",
        default=config.assets_root,
    )
    parser.add_argument(
        "--skip_default_style",
        action="store_true",
        help="Skip saving default style config and mean vector.",
    )
    parser.add_argument(
        "--no_progress_bar",
        action="store_true",
        help="Do not show the progress bar while training.",
    )
    parser.add_argument(
        "--speedup",
        action="store_true",
        help="Speed up training by disabling logging and evaluation.",
    )
    args = parser.parse_args()

    # Set log file
    model_dir = os.path.join(args.model, config.train_ms_config.model_dir)
    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
    logger.add(os.path.join(args.model, f"train_{timestamp}.log"))

    # Parsing environment variables
    envs = config.train_ms_config.env
    for env_name, env_value in envs.items():
        if env_name not in os.environ.keys():
            logger.info("Loading configuration from config {}".format(str(env_value)))
            os.environ[env_name] = str(env_value)
    logger.info(
        "Loading environment variables \nMASTER_ADDR: {},\nMASTER_PORT: {},\nWORLD_SIZE: {},\nRANK: {},\nLOCAL_RANK: {}".format(
            os.environ["MASTER_ADDR"],
            os.environ["MASTER_PORT"],
            os.environ["WORLD_SIZE"],
            os.environ["RANK"],
            os.environ["LOCAL_RANK"],
        )
    )

    backend = "nccl"
    if platform.system() == "Windows":
        backend = "gloo"  # If Windows,switch to gloo backend.
    dist.init_process_group(
        backend=backend,
        init_method="env://",
        timeout=datetime.timedelta(seconds=300),
    )  # Use torchrun instead of mp.spawn
    rank = dist.get_rank()
    local_rank = int(os.environ["LOCAL_RANK"])
    n_gpus = dist.get_world_size()

    hps = utils.get_hparams_from_file(args.config)
    # This is needed because we have to pass values to `train_and_evaluate()
    hps.model_dir = model_dir
    hps.speedup = args.speedup

    # 比较路径是否相同
    if os.path.realpath(args.config) != os.path.realpath(
        config.train_ms_config.config_path
    ):
        with open(args.config, "r", encoding="utf-8") as f:
            data = f.read()
        os.makedirs(os.path.dirname(config.train_ms_config.config_path), exist_ok=True)
        with open(config.train_ms_config.config_path, "w", encoding="utf-8") as f:
            f.write(data)

    """
    Path constants are a bit complicated...
    TODO: Refactor or rename these?
    (Both `config.yml` and `config.json` are used, which is confusing I think.)

    args.model: For saving all info needed for training.
        default: `Data/{model_name}`.
    hps.model_dir := model_dir: For saving checkpoints (for resuming training).
        default: `Data/{model_name}/models`.
        (Use `hps` since we have to pass `model_dir` to `train_and_evaluate()`.

    args.assets_root: The root directory of model assets needed for inference.
        default: config.assets_root == `model_assets`.

    config.out_dir: The directory for model assets of this model (for inference).
        default: `model_assets/{model_name}`.
    """
    os.makedirs(config.out_dir, exist_ok=True)

    if not args.skip_default_style:
        # Save default style to out_dir
        default_style.set_style_config(
            args.config, os.path.join(config.out_dir, "config.json")
        )
        default_style.save_mean_vector(
            os.path.join(args.model, "wavs"),
            os.path.join(config.out_dir, "style_vectors.npy"),
        )

    torch.manual_seed(hps.train.seed)
    torch.cuda.set_device(local_rank)

    global global_step
    writer = None
    writer_eval = None
    if rank == 0 and not args.speedup:
        # logger = utils.get_logger(hps.model_dir)
        # logger.info(hps)
        utils.check_git_hash(model_dir)
        writer = SummaryWriter(log_dir=model_dir)
        writer_eval = SummaryWriter(log_dir=os.path.join(model_dir, "eval"))
    train_dataset = TextAudioSpeakerLoader(hps.data.training_files, hps.data)
    train_sampler = DistributedBucketSampler(
        train_dataset,
        hps.train.batch_size,
        [32, 300, 400, 500, 600, 700, 800, 900, 1000],
        num_replicas=n_gpus,
        rank=rank,
        shuffle=True,
    )
    collate_fn = TextAudioSpeakerCollate(use_jp_extra=True)
    train_loader = DataLoader(
        train_dataset,
        # メモリ消費量を減らそうとnum_workersを1にしてみる
        # num_workers=min(config.train_ms_config.num_workers, os.cpu_count() // 2),
        num_workers=1,
        shuffle=False,
        pin_memory=True,
        collate_fn=collate_fn,
        batch_sampler=train_sampler,
        persistent_workers=True,
        # これもメモリ消費量を減らそうとしてコメントアウト
        # prefetch_factor=6,
    )  # DataLoader config could be adjusted.
    eval_dataset = None
    eval_loader = None
    if rank == 0 and not args.speedup:
        eval_dataset = TextAudioSpeakerLoader(hps.data.validation_files, hps.data)
        eval_loader = DataLoader(
            eval_dataset,
            num_workers=0,
            shuffle=False,
            batch_size=1,
            pin_memory=True,
            drop_last=False,
            collate_fn=collate_fn,
        )
    if (
        "use_noise_scaled_mas" in hps.model.keys()
        and hps.model.use_noise_scaled_mas is True
    ):
        logger.info("Using noise scaled MAS for VITS2")
        mas_noise_scale_initial = 0.01
        noise_scale_delta = 2e-6
    else:
        logger.info("Using normal MAS for VITS1")
        mas_noise_scale_initial = 0.0
        noise_scale_delta = 0.0
    if (
        "use_duration_discriminator" in hps.model.keys()
        and hps.model.use_duration_discriminator is True
    ):
        logger.info("Using duration discriminator for VITS2")
        net_dur_disc = DurationDiscriminator(
            hps.model.hidden_channels,
            hps.model.hidden_channels,
            3,
            0.1,
            gin_channels=hps.model.gin_channels if hps.data.n_speakers != 0 else 0,
        ).cuda(local_rank)
    else:
        net_dur_disc = None
    if (
        "use_wavlm_discriminator" in hps.model.keys()
        and hps.model.use_wavlm_discriminator is True
    ):
        net_wd = WavLMDiscriminator(
            hps.model.slm.hidden, hps.model.slm.nlayers, hps.model.slm.initial_channel
        ).cuda(local_rank)
    else:
        net_wd = None
    if (
        "use_spk_conditioned_encoder" in hps.model.keys()
        and hps.model.use_spk_conditioned_encoder is True
    ):
        if hps.data.n_speakers == 0:
            raise ValueError(
                "n_speakers must be > 0 when using spk conditioned encoder to train multi-speaker model"
            )
    else:
        logger.info("Using normal encoder for VITS1")

    net_g = SynthesizerTrn(
        len(symbols),
        hps.data.filter_length // 2 + 1,
        hps.train.segment_size // hps.data.hop_length,
        n_speakers=hps.data.n_speakers,
        mas_noise_scale_initial=mas_noise_scale_initial,
        noise_scale_delta=noise_scale_delta,
        **hps.model,
    ).cuda(local_rank)
    if getattr(hps.train, "freeze_JP_bert", False):
        logger.info("Freezing (JP) bert encoder !!!")
        for param in net_g.enc_p.bert_proj.parameters():
            param.requires_grad = False
    if getattr(hps.train, "freeze_style", False):
        logger.info("Freezing style encoder !!!")
        for param in net_g.enc_p.style_proj.parameters():
            param.requires_grad = False

    net_d = MultiPeriodDiscriminator(hps.model.use_spectral_norm).cuda(local_rank)
    optim_g = torch.optim.AdamW(
        filter(lambda p: p.requires_grad, net_g.parameters()),
        hps.train.learning_rate,
        betas=hps.train.betas,
        eps=hps.train.eps,
    )
    optim_d = torch.optim.AdamW(
        net_d.parameters(),
        hps.train.learning_rate,
        betas=hps.train.betas,
        eps=hps.train.eps,
    )
    if net_dur_disc is not None:
        optim_dur_disc = torch.optim.AdamW(
            net_dur_disc.parameters(),
            hps.train.learning_rate,
            betas=hps.train.betas,
            eps=hps.train.eps,
        )
    else:
        optim_dur_disc = None
    if net_wd is not None:
        optim_wd = torch.optim.AdamW(
            net_wd.parameters(),
            hps.train.learning_rate,
            betas=hps.train.betas,
            eps=hps.train.eps,
        )
    else:
        optim_wd = None
    net_g = DDP(
        net_g,
        device_ids=[local_rank],
        # bucket_cap_mb=512
    )
    net_d = DDP(
        net_d,
        device_ids=[local_rank],
        # bucket_cap_mb=512
    )
    if net_dur_disc is not None:
        net_dur_disc = DDP(
            net_dur_disc,
            device_ids=[local_rank],
            # bucket_cap_mb=512,
        )
    if net_wd is not None:
        net_wd = DDP(
            net_wd,
            device_ids=[local_rank],
            #  bucket_cap_mb=512
        )

    if utils.is_resuming(model_dir):
        if net_dur_disc is not None:
            try:
                _, _, dur_resume_lr, epoch_str = utils.load_checkpoint(
                    utils.latest_checkpoint_path(model_dir, "DUR_*.pth"),
                    net_dur_disc,
                    optim_dur_disc,
                    skip_optimizer=(
                        hps.train.skip_optimizer
                        if "skip_optimizer" in hps.train
                        else True
                    ),
                )
                if not optim_dur_disc.param_groups[0].get("initial_lr"):
                    optim_dur_disc.param_groups[0]["initial_lr"] = dur_resume_lr
            except:
                if not optim_dur_disc.param_groups[0].get("initial_lr"):
                    optim_dur_disc.param_groups[0]["initial_lr"] = dur_resume_lr
                print("Initialize dur_disc")
        if net_wd is not None:
            try:
                _, optim_wd, wd_resume_lr, epoch_str = utils.load_checkpoint(
                    utils.latest_checkpoint_path(model_dir, "WD_*.pth"),
                    net_wd,
                    optim_wd,
                    skip_optimizer=(
                        hps.train.skip_optimizer
                        if "skip_optimizer" in hps.train
                        else True
                    ),
                )
                if not optim_wd.param_groups[0].get("initial_lr"):
                    optim_wd.param_groups[0]["initial_lr"] = wd_resume_lr
            except:
                if not optim_wd.param_groups[0].get("initial_lr"):
                    optim_wd.param_groups[0]["initial_lr"] = wd_resume_lr
                logger.info("Initialize wavlm")

        try:
            _, optim_g, g_resume_lr, epoch_str = utils.load_checkpoint(
                utils.latest_checkpoint_path(model_dir, "G_*.pth"),
                net_g,
                optim_g,
                skip_optimizer=(
                    hps.train.skip_optimizer if "skip_optimizer" in hps.train else True
                ),
            )
            _, optim_d, d_resume_lr, epoch_str = utils.load_checkpoint(
                utils.latest_checkpoint_path(model_dir, "D_*.pth"),
                net_d,
                optim_d,
                skip_optimizer=(
                    hps.train.skip_optimizer if "skip_optimizer" in hps.train else True
                ),
            )
            if not optim_g.param_groups[0].get("initial_lr"):
                optim_g.param_groups[0]["initial_lr"] = g_resume_lr
            if not optim_d.param_groups[0].get("initial_lr"):
                optim_d.param_groups[0]["initial_lr"] = d_resume_lr

            epoch_str = max(epoch_str, 1)
            # global_step = (epoch_str - 1) * len(train_loader)
            global_step = int(
                utils.get_steps(utils.latest_checkpoint_path(model_dir, "G_*.pth"))
            )
            logger.info(
                f"******************Found the model. Current epoch is {epoch_str}, gloabl step is {global_step}*********************"
            )
        except Exception as e:
            logger.warning(e)
            logger.warning(
                "It seems that you are not using the pretrained models, so we will train from scratch."
            )
            epoch_str = 1
            global_step = 0
    else:
        try:
            _ = utils.load_safetensors(
                os.path.join(model_dir, "G_0.safetensors"), net_g
            )
            _ = utils.load_safetensors(
                os.path.join(model_dir, "D_0.safetensors"), net_d
            )
            if net_dur_disc is not None:
                _ = utils.load_safetensors(
                    os.path.join(model_dir, "DUR_0.safetensors"), net_dur_disc
                )
            if net_wd is not None:
                _ = utils.load_safetensors(
                    os.path.join(model_dir, "WD_0.safetensors"), net_wd
                )
            logger.info("Loaded the pretrained models.")
        except Exception as e:
            logger.warning(e)
            logger.warning(
                "It seems that you are not using the pretrained models, so we will train from scratch."
            )
        finally:
            epoch_str = 1
            global_step = 0

    def lr_lambda(epoch):
        """
        Learning rate scheduler for warmup and exponential decay.
        - During the warmup period, the learning rate increases linearly.
        - After the warmup period, the learning rate decreases exponentially.
        """
        if epoch < hps.train.warmup_epochs:
            return float(epoch) / float(max(1, hps.train.warmup_epochs))
        else:
            return hps.train.lr_decay ** (epoch - hps.train.warmup_epochs)

    scheduler_last_epoch = epoch_str - 2
    scheduler_g = torch.optim.lr_scheduler.LambdaLR(
        optim_g, lr_lambda=lr_lambda, last_epoch=scheduler_last_epoch
    )
    scheduler_d = torch.optim.lr_scheduler.LambdaLR(
        optim_d, lr_lambda=lr_lambda, last_epoch=scheduler_last_epoch
    )
    if net_dur_disc is not None:
        scheduler_dur_disc = torch.optim.lr_scheduler.LambdaLR(
            optim_dur_disc, lr_lambda=lr_lambda, last_epoch=scheduler_last_epoch
        )
    else:
        scheduler_dur_disc = None
    if net_wd is not None:
        scheduler_wd = torch.optim.lr_scheduler.LambdaLR(
            optim_wd, lr_lambda=lr_lambda, last_epoch=scheduler_last_epoch
        )
        wl = WavLMLoss(
            hps.model.slm.model,
            net_wd,
            hps.data.sampling_rate,
            hps.model.slm.sr,
        ).to(local_rank)
    else:
        scheduler_wd = None
        wl = None
    scaler = GradScaler(enabled=hps.train.bf16_run)
    logger.info("Start training.")

    diff = abs(
        epoch_str * len(train_loader) - (hps.train.epochs + 1) * len(train_loader)
    )
    pbar = None
    if not args.no_progress_bar:
        pbar = tqdm(
            total=global_step + diff,
            initial=global_step,
            smoothing=0.05,
            file=SAFE_STDOUT,
        )
    initial_step = global_step

    for epoch in range(epoch_str, hps.train.epochs + 1):
        if rank == 0:
            train_and_evaluate(
                rank,
                local_rank,
                epoch,
                hps,
                [net_g, net_d, net_dur_disc, net_wd, wl],
                [optim_g, optim_d, optim_dur_disc, optim_wd],
                [scheduler_g, scheduler_d, scheduler_dur_disc, scheduler_wd],
                scaler,
                [train_loader, eval_loader],
                logger,
                [writer, writer_eval],
                pbar,
                initial_step,
            )
        else:
            train_and_evaluate(
                rank,
                local_rank,
                epoch,
                hps,
                [net_g, net_d, net_dur_disc, net_wd, wl],
                [optim_g, optim_d, optim_dur_disc, optim_wd],
                [scheduler_g, scheduler_d, scheduler_dur_disc, scheduler_wd],
                scaler,
                [train_loader, None],
                None,
                None,
                pbar,
                initial_step,
            )
        scheduler_g.step()
        scheduler_d.step()
        if net_dur_disc is not None:
            scheduler_dur_disc.step()
        if net_wd is not None:
            scheduler_wd.step()
        if epoch == hps.train.epochs:
            # Save the final models
            utils.save_checkpoint(
                net_g,
                optim_g,
                hps.train.learning_rate,
                epoch,
                os.path.join(model_dir, "G_{}.pth".format(global_step)),
            )
            utils.save_checkpoint(
                net_d,
                optim_d,
                hps.train.learning_rate,
                epoch,
                os.path.join(model_dir, "D_{}.pth".format(global_step)),
            )
            if net_dur_disc is not None:
                utils.save_checkpoint(
                    net_dur_disc,
                    optim_dur_disc,
                    hps.train.learning_rate,
                    epoch,
                    os.path.join(model_dir, "DUR_{}.pth".format(global_step)),
                )
            if net_wd is not None:
                utils.save_checkpoint(
                    net_wd,
                    optim_wd,
                    hps.train.learning_rate,
                    epoch,
                    os.path.join(model_dir, "WD_{}.pth".format(global_step)),
                )
            utils.save_safetensors(
                net_g,
                epoch,
                os.path.join(
                    config.out_dir,
                    f"{config.model_name}_e{epoch}_s{global_step}.safetensors",
                ),
                for_infer=True,
            )

    if pbar is not None:
        pbar.close()


def train_and_evaluate(
    rank,
    local_rank,
    epoch,
    hps,
    nets,
    optims,
    schedulers,
    scaler,
    loaders,
    logger,
    writers,
    pbar: tqdm,
    initial_step: int,
):
    net_g, net_d, net_dur_disc, net_wd, wl = nets
    optim_g, optim_d, optim_dur_disc, optim_wd = optims
    scheduler_g, scheduler_d, scheduler_dur_disc, scheduler_wd = schedulers
    train_loader, eval_loader = loaders
    if writers is not None:
        writer, writer_eval = writers

    train_loader.batch_sampler.set_epoch(epoch)
    global global_step

    net_g.train()
    net_d.train()
    if net_dur_disc is not None:
        net_dur_disc.train()
    if net_wd is not None:
        net_wd.train()
    for batch_idx, (
        x,
        x_lengths,
        spec,
        spec_lengths,
        y,
        y_lengths,
        speakers,
        tone,
        language,
        bert,
        style_vec,
    ) in enumerate(train_loader):
        if net_g.module.use_noise_scaled_mas:
            current_mas_noise_scale = (
                net_g.module.mas_noise_scale_initial
                - net_g.module.noise_scale_delta * global_step
            )
            net_g.module.current_mas_noise_scale = max(current_mas_noise_scale, 0.0)
        x, x_lengths = x.cuda(local_rank, non_blocking=True), x_lengths.cuda(
            local_rank, non_blocking=True
        )
        spec, spec_lengths = spec.cuda(
            local_rank, non_blocking=True
        ), spec_lengths.cuda(local_rank, non_blocking=True)
        y, y_lengths = y.cuda(local_rank, non_blocking=True), y_lengths.cuda(
            local_rank, non_blocking=True
        )
        speakers = speakers.cuda(local_rank, non_blocking=True)
        tone = tone.cuda(local_rank, non_blocking=True)
        language = language.cuda(local_rank, non_blocking=True)
        bert = bert.cuda(local_rank, non_blocking=True)
        style_vec = style_vec.cuda(local_rank, non_blocking=True)

        with autocast(enabled=hps.train.bf16_run, dtype=torch.bfloat16):
            (
                y_hat,
                l_length,
                attn,
                ids_slice,
                x_mask,
                z_mask,
                (z, z_p, m_p, logs_p, m_q, logs_q),
                (hidden_x, logw, logw_),  # , logw_sdp),
                g,
            ) = net_g(
                x,
                x_lengths,
                spec,
                spec_lengths,
                speakers,
                tone,
                language,
                bert,
                style_vec,
            )
            mel = spec_to_mel_torch(
                spec,
                hps.data.filter_length,
                hps.data.n_mel_channels,
                hps.data.sampling_rate,
                hps.data.mel_fmin,
                hps.data.mel_fmax,
            )
            y_mel = commons.slice_segments(
                mel, ids_slice, hps.train.segment_size // hps.data.hop_length
            )
            y_hat_mel = mel_spectrogram_torch(
                y_hat.squeeze(1).float(),
                hps.data.filter_length,
                hps.data.n_mel_channels,
                hps.data.sampling_rate,
                hps.data.hop_length,
                hps.data.win_length,
                hps.data.mel_fmin,
                hps.data.mel_fmax,
            )

            y = commons.slice_segments(
                y, ids_slice * hps.data.hop_length, hps.train.segment_size
            )  # slice

            # Discriminator
            y_d_hat_r, y_d_hat_g, _, _ = net_d(y, y_hat.detach())
            with autocast(enabled=hps.train.bf16_run, dtype=torch.bfloat16):
                loss_disc, losses_disc_r, losses_disc_g = discriminator_loss(
                    y_d_hat_r, y_d_hat_g
                )
                loss_disc_all = loss_disc
            if net_dur_disc is not None:
                y_dur_hat_r, y_dur_hat_g = net_dur_disc(
                    hidden_x.detach(),
                    x_mask.detach(),
                    logw_.detach(),
                    logw.detach(),
                    g.detach(),
                )
                with autocast(enabled=hps.train.bf16_run, dtype=torch.bfloat16):
                    # TODO: I think need to mean using the mask, but for now, just mean all
                    (
                        loss_dur_disc,
                        losses_dur_disc_r,
                        losses_dur_disc_g,
                    ) = discriminator_loss(y_dur_hat_r, y_dur_hat_g)
                    loss_dur_disc_all = loss_dur_disc
                optim_dur_disc.zero_grad()
                scaler.scale(loss_dur_disc_all).backward()
                scaler.unscale_(optim_dur_disc)
                # torch.nn.utils.clip_grad_norm_(
                # parameters=net_dur_disc.parameters(), max_norm=5
                # )
                grad_norm_dur = commons.clip_grad_value_(
                    net_dur_disc.parameters(), None
                )
                scaler.step(optim_dur_disc)
            if net_wd is not None:
                # logger.debug(f"y.shape: {y.shape}, y_hat.shape: {y_hat.shape}")
                # shape: (batch, 1, time)
                with autocast(enabled=hps.train.bf16_run, dtype=torch.bfloat16):
                    loss_slm = wl.discriminator(
                        y.detach().squeeze(1), y_hat.detach().squeeze(1)
                    ).mean()

                optim_wd.zero_grad()
                scaler.scale(loss_slm).backward()
                scaler.unscale_(optim_wd)
                # torch.nn.utils.clip_grad_norm_(parameters=net_wd.parameters(), max_norm=200)
                grad_norm_wd = commons.clip_grad_value_(net_wd.parameters(), None)
                scaler.step(optim_wd)

        optim_d.zero_grad()
        scaler.scale(loss_disc_all).backward()
        scaler.unscale_(optim_d)
        if getattr(hps.train, "bf16_run", False):
            torch.nn.utils.clip_grad_norm_(parameters=net_d.parameters(), max_norm=200)
        grad_norm_d = commons.clip_grad_value_(net_d.parameters(), None)
        scaler.step(optim_d)

        with autocast(enabled=hps.train.bf16_run, dtype=torch.bfloat16):
            # Generator
            y_d_hat_r, y_d_hat_g, fmap_r, fmap_g = net_d(y, y_hat)
            if net_dur_disc is not None:
                _, y_dur_hat_g = net_dur_disc(hidden_x, x_mask, logw_, logw, g)
            if net_wd is not None:
                loss_lm = wl(y.detach().squeeze(1), y_hat.squeeze(1)).mean()
                loss_lm_gen = wl.generator(y_hat.squeeze(1))
            with autocast(enabled=hps.train.bf16_run, dtype=torch.bfloat16):
                loss_dur = torch.sum(l_length.float())
                loss_mel = F.l1_loss(y_mel, y_hat_mel) * hps.train.c_mel
                loss_kl = kl_loss(z_p, logs_q, m_p, logs_p, z_mask) * hps.train.c_kl

                loss_fm = feature_loss(fmap_r, fmap_g)
                loss_gen, losses_gen = generator_loss(y_d_hat_g)
                # loss_commit = loss_commit * hps.train.c_commit

                loss_gen_all = loss_gen + loss_fm + loss_mel + loss_dur + loss_kl
                if net_dur_disc is not None:
                    loss_dur_gen, losses_dur_gen = generator_loss(y_dur_hat_g)
                    if net_wd is not None:
                        loss_gen_all += loss_dur_gen + loss_lm + loss_lm_gen
                    else:
                        loss_gen_all += loss_dur_gen
        optim_g.zero_grad()
        scaler.scale(loss_gen_all).backward()
        scaler.unscale_(optim_g)
        # if getattr(hps.train, "bf16_run", False):
        torch.nn.utils.clip_grad_norm_(parameters=net_g.parameters(), max_norm=500)
        grad_norm_g = commons.clip_grad_value_(net_g.parameters(), None)
        scaler.step(optim_g)
        scaler.update()

        if rank == 0:
            if global_step % hps.train.log_interval == 0 and not hps.speedup:
                lr = optim_g.param_groups[0]["lr"]
                losses = [loss_disc, loss_gen, loss_fm, loss_mel, loss_dur, loss_kl]
                # logger.info(
                #     "Train Epoch: {} [{:.0f}%]".format(
                #         epoch, 100.0 * batch_idx / len(train_loader)
                #     )
                # )
                # logger.info([x.item() for x in losses] + [global_step, lr])

                scalar_dict = {
                    "loss/g/total": loss_gen_all,
                    "loss/d/total": loss_disc_all,
                    "learning_rate": lr,
                    "grad_norm_d": grad_norm_d,
                    "grad_norm_g": grad_norm_g,
                }
                scalar_dict.update(
                    {
                        "loss/g/fm": loss_fm,
                        "loss/g/mel": loss_mel,
                        "loss/g/dur": loss_dur,
                        "loss/g/kl": loss_kl,
                    }
                )
                scalar_dict.update(
                    {"loss/g/{}".format(i): v for i, v in enumerate(losses_gen)}
                )
                scalar_dict.update(
                    {"loss/d_r/{}".format(i): v for i, v in enumerate(losses_disc_r)}
                )
                scalar_dict.update(
                    {"loss/d_g/{}".format(i): v for i, v in enumerate(losses_disc_g)}
                )

                if net_dur_disc is not None:
                    scalar_dict.update({"loss/dur_disc/total": loss_dur_disc_all})

                    scalar_dict.update(
                        {
                            "loss/dur_disc_g/{}".format(i): v
                            for i, v in enumerate(losses_dur_disc_g)
                        }
                    )
                    scalar_dict.update(
                        {
                            "loss/dur_disc_r/{}".format(i): v
                            for i, v in enumerate(losses_dur_disc_r)
                        }
                    )

                    scalar_dict.update({"loss/g/dur_gen": loss_dur_gen})
                    scalar_dict.update(
                        {
                            "loss/g/dur_gen_{}".format(i): v
                            for i, v in enumerate(losses_dur_gen)
                        }
                    )

                if net_wd is not None:
                    scalar_dict.update(
                        {
                            "loss/wd/total": loss_slm,
                            "grad_norm_wd": grad_norm_wd,
                            "loss/g/lm": loss_lm,
                            "loss/g/lm_gen": loss_lm_gen,
                        }
                    )
                image_dict = {
                    "slice/mel_org": utils.plot_spectrogram_to_numpy(
                        y_mel[0].data.cpu().numpy()
                    ),
                    "slice/mel_gen": utils.plot_spectrogram_to_numpy(
                        y_hat_mel[0].data.cpu().numpy()
                    ),
                    "all/mel": utils.plot_spectrogram_to_numpy(
                        mel[0].data.cpu().numpy()
                    ),
                    "all/attn": utils.plot_alignment_to_numpy(
                        attn[0, 0].data.cpu().numpy()
                    ),
                }
                utils.summarize(
                    writer=writer,
                    global_step=global_step,
                    images=image_dict,
                    scalars=scalar_dict,
                )

            if (
                global_step % hps.train.eval_interval == 0
                and global_step != 0
                and initial_step != global_step
            ):
                if not hps.speedup:
                    evaluate(hps, net_g, eval_loader, writer_eval)
                utils.save_checkpoint(
                    net_g,
                    optim_g,
                    hps.train.learning_rate,
                    epoch,
                    os.path.join(hps.model_dir, "G_{}.pth".format(global_step)),
                )
                utils.save_checkpoint(
                    net_d,
                    optim_d,
                    hps.train.learning_rate,
                    epoch,
                    os.path.join(hps.model_dir, "D_{}.pth".format(global_step)),
                )
                if net_dur_disc is not None:
                    utils.save_checkpoint(
                        net_dur_disc,
                        optim_dur_disc,
                        hps.train.learning_rate,
                        epoch,
                        os.path.join(hps.model_dir, "DUR_{}.pth".format(global_step)),
                    )
                if net_wd is not None:
                    utils.save_checkpoint(
                        net_wd,
                        optim_wd,
                        hps.train.learning_rate,
                        epoch,
                        os.path.join(hps.model_dir, "WD_{}.pth".format(global_step)),
                    )
                keep_ckpts = config.train_ms_config.keep_ckpts
                if keep_ckpts > 0:
                    utils.clean_checkpoints(
                        path_to_models=hps.model_dir,
                        n_ckpts_to_keep=keep_ckpts,
                        sort_by_time=True,
                    )
                # Save safetensors (for inference) to `model_assets/{model_name}`
                utils.save_safetensors(
                    net_g,
                    epoch,
                    os.path.join(
                        config.out_dir,
                        f"{config.model_name}_e{epoch}_s{global_step}.safetensors",
                    ),
                    for_infer=True,
                )

        global_step += 1
        if pbar is not None:
            pbar.set_description(
                "Epoch {}({:.0f}%)/{}".format(
                    epoch, 100.0 * batch_idx / len(train_loader), hps.train.epochs
                )
            )
            pbar.update()
    # 本家ではこれをスピードアップのために消すと書かれていたので、一応消してみる
    # と思ったけどメモリ使用量が減るかもしれないのでつけてみる
    gc.collect()
    torch.cuda.empty_cache()
    if pbar is None and rank == 0:
        logger.info(f"====> Epoch: {epoch}, step: {global_step}")


def evaluate(hps, generator, eval_loader, writer_eval):
    generator.eval()
    image_dict = {}
    audio_dict = {}
    logger.info("Evaluating ...")
    with torch.no_grad():
        for batch_idx, (
            x,
            x_lengths,
            spec,
            spec_lengths,
            y,
            y_lengths,
            speakers,
            tone,
            language,
            bert,
            style_vec,
        ) in enumerate(eval_loader):
            x, x_lengths = x.cuda(), x_lengths.cuda()
            spec, spec_lengths = spec.cuda(), spec_lengths.cuda()
            y, y_lengths = y.cuda(), y_lengths.cuda()
            speakers = speakers.cuda()
            bert = bert.cuda()
            tone = tone.cuda()
            language = language.cuda()
            style_vec = style_vec.cuda()
            for use_sdp in [True, False]:
                y_hat, attn, mask, *_ = generator.module.infer(
                    x,
                    x_lengths,
                    speakers,
                    tone,
                    language,
                    bert,
                    style_vec,
                    y=spec,
                    max_len=1000,
                    sdp_ratio=0.0 if not use_sdp else 1.0,
                )
                y_hat_lengths = mask.sum([1, 2]).long() * hps.data.hop_length

                mel = spec_to_mel_torch(
                    spec,
                    hps.data.filter_length,
                    hps.data.n_mel_channels,
                    hps.data.sampling_rate,
                    hps.data.mel_fmin,
                    hps.data.mel_fmax,
                )
                y_hat_mel = mel_spectrogram_torch(
                    y_hat.squeeze(1).float(),
                    hps.data.filter_length,
                    hps.data.n_mel_channels,
                    hps.data.sampling_rate,
                    hps.data.hop_length,
                    hps.data.win_length,
                    hps.data.mel_fmin,
                    hps.data.mel_fmax,
                )
                image_dict.update(
                    {
                        f"gen/mel_{batch_idx}": utils.plot_spectrogram_to_numpy(
                            y_hat_mel[0].cpu().numpy()
                        )
                    }
                )
                audio_dict.update(
                    {
                        f"gen/audio_{batch_idx}_{use_sdp}": y_hat[
                            0, :, : y_hat_lengths[0]
                        ]
                    }
                )
                image_dict.update(
                    {
                        f"gt/mel_{batch_idx}": utils.plot_spectrogram_to_numpy(
                            mel[0].cpu().numpy()
                        )
                    }
                )
                audio_dict.update({f"gt/audio_{batch_idx}": y[0, :, : y_lengths[0]]})

    utils.summarize(
        writer=writer_eval,
        global_step=global_step,
        images=image_dict,
        audios=audio_dict,
        audio_sampling_rate=hps.data.sampling_rate,
    )
    generator.train()


if __name__ == "__main__":
    run()