#!/usr/bin/env python3 # Copyright 2023 Xiaomi Corp. (Author: Zengwei Yao) # 2024 The Chinese University of HK (Author: Zengrui Jin) # # See ../../../../LICENSE for clarification regarding multiple authors # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import argparse import itertools import logging import math import random from pathlib import Path from shutil import copyfile from typing import Any, Dict, Optional, Tuple, Union import numpy as np import torch import torch.multiprocessing as mp import torch.nn as nn from codec_datamodule import LibriTTSCodecDataModule from encodec import Encodec from lhotse.utils import fix_random_seed from scheduler import WarmupCosineLrScheduler from torch import nn from torch.cuda.amp import GradScaler, autocast from torch.nn.parallel import DistributedDataParallel as DDP from torch.optim import Optimizer from torch.utils.tensorboard import SummaryWriter from utils import MetricsTracker, save_checkpoint from icefall import diagnostics from icefall.checkpoint import load_checkpoint from icefall.dist import cleanup_dist, setup_dist from icefall.env import get_env_info from icefall.hooks import register_inf_check_hooks from icefall.utils import AttributeDict, setup_logger, str2bool LRSchedulerType = torch.optim.lr_scheduler._LRScheduler def get_parser(): parser = argparse.ArgumentParser( formatter_class=argparse.ArgumentDefaultsHelpFormatter ) parser.add_argument( "--world-size", type=int, default=1, help="Number of GPUs for DDP training.", ) parser.add_argument( "--master-port", type=int, default=12354, help="Master port to use for DDP training.", ) parser.add_argument( "--tensorboard", type=str2bool, default=True, help="Should various information be logged in tensorboard.", ) parser.add_argument( "--num-samples", type=int, default=3, help="Number of samples to generate for tensorboard.", ) parser.add_argument( "--num-epochs", type=int, default=500, help="Number of epochs to train.", ) parser.add_argument( "--start-epoch", type=int, default=1, help="""Resume training from this epoch. It should be positive. If larger than 1, it will load checkpoint from exp-dir/epoch-{start_epoch-1}.pt """, ) parser.add_argument( "--exp-dir", type=str, default="encodec/exp", help="""The experiment dir. It specifies the directory where all training related files, e.g., checkpoints, log, etc, are saved """, ) parser.add_argument( "--lr", type=float, default=3.0e-4, help="The base learning rate." ) parser.add_argument( "--seed", type=int, default=42, help="The seed for random generators intended for reproducibility", ) parser.add_argument( "--print-diagnostics", type=str2bool, default=False, help="Accumulate stats on activations, print them and exit.", ) parser.add_argument( "--inf-check", type=str2bool, default=False, help="Add hooks to check for infinite module outputs and gradients.", ) parser.add_argument( "--save-every-n", type=int, default=5, help="""Save checkpoint after processing this number of epochs" periodically. We save checkpoint to exp-dir/ whenever params.cur_epoch % save_every_n == 0. The checkpoint filename has the form: f'exp-dir/epoch-{params.cur_epoch}.pt'. Since it will take around 1000 epochs, we suggest using a large save_every_n to save disk space. """, ) parser.add_argument( "--use-fp16", type=str2bool, default=False, help="Whether to use half precision training.", ) return parser def get_params() -> AttributeDict: """Return a dict containing training parameters. All training related parameters that are not passed from the commandline are saved in the variable `params`. Commandline options are merged into `params` after they are parsed, so you can also access them via `params`. Explanation of options saved in `params`: - best_train_loss: Best training loss so far. It is used to select the model that has the lowest training loss. It is updated during the training. - best_valid_loss: Best validation loss so far. It is used to select the model that has the lowest validation loss. It is updated during the training. - best_train_epoch: It is the epoch that has the best training loss. - best_valid_epoch: It is the epoch that has the best validation loss. - batch_idx_train: Used to writing statistics to tensorboard. It contains number of batches trained so far across epochs. - log_interval: Print training loss if batch_idx % log_interval` is 0 - valid_interval: Run validation if batch_idx % valid_interval is 0 """ params = AttributeDict( { # training params "best_train_loss": float("inf"), "best_valid_loss": float("inf"), "best_train_epoch": -1, "best_valid_epoch": -1, "batch_idx_train": -1, # 0 "log_interval": 50, "valid_interval": 200, "env_info": get_env_info(), "sampling_rate": 24000, "audio_normalization": False, "lambda_adv": 3.0, # loss scaling coefficient for adversarial loss "lambda_wav": 0.1, # loss scaling coefficient for waveform loss "lambda_feat": 4.0, # loss scaling coefficient for feat loss "lambda_rec": 1.0, # loss scaling coefficient for reconstruction loss "lambda_com": 1000.0, # loss scaling coefficient for commitment loss } ) return params def load_checkpoint_if_available( params: AttributeDict, model: nn.Module ) -> Optional[Dict[str, Any]]: """Load checkpoint from file. If params.start_epoch is larger than 1, it will load the checkpoint from `params.start_epoch - 1`. Apart from loading state dict for `model` and `optimizer` it also updates `best_train_epoch`, `best_train_loss`, `best_valid_epoch`, and `best_valid_loss` in `params`. Args: params: The return value of :func:`get_params`. model: The training model. Returns: Return a dict containing previously saved training info. """ if params.start_epoch > 1: filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt" else: return None assert filename.is_file(), f"{filename} does not exist!" saved_params = load_checkpoint(filename, model=model) keys = [ "best_train_epoch", "best_valid_epoch", "batch_idx_train", "best_train_loss", "best_valid_loss", ] for k in keys: params[k] = saved_params[k] return saved_params def get_model(params: AttributeDict) -> nn.Module: """Get the model based on the configuration.""" from discriminators import ( MultiPeriodDiscriminator, MultiScaleDiscriminator, MultiScaleSTFTDiscriminator, ) from modules.seanet import SEANetDecoder, SEANetEncoder from quantization import ResidualVectorQuantizer # generator_params = { # "generator_n_filters": 32, # "dimension": 512, # "ratios": [2, 2, 2, 4], # "target_bandwidths": [7.5, 15], # "bins": 1024, # } # discriminator_params = { # "stft_discriminator_n_filters": 32, # "discriminator_epoch_start": 5, # } # inference_params = { # "target_bw": 7.5, # } generator_params = { "generator_n_filters": 32, "dimension": 512, "ratios": [8, 5, 4, 2], "target_bandwidths": [1.5, 3, 6, 12, 24], "bins": 1024, } discriminator_params = { "stft_discriminator_n_filters": 32, "discriminator_epoch_start": 5, "n_ffts": [1024, 2048, 512], "hop_lengths": [256, 512, 128], "win_lengths": [1024, 2048, 512], } inference_params = { "target_bw": 6, } params.update(generator_params) params.update(discriminator_params) params.update(inference_params) hop_length = np.prod(params.ratios) n_q = int( 1000 * params.target_bandwidths[-1] // (math.ceil(params.sampling_rate / hop_length) * 10) ) encoder = SEANetEncoder( n_filters=params.generator_n_filters, dimension=params.dimension, ratios=params.ratios, ) decoder = SEANetDecoder( n_filters=params.generator_n_filters, dimension=params.dimension, ratios=params.ratios, ) quantizer = ResidualVectorQuantizer( dimension=params.dimension, n_q=n_q, bins=params.bins ) model = Encodec( params=params, sampling_rate=params.sampling_rate, target_bandwidths=params.target_bandwidths, encoder=encoder, quantizer=quantizer, decoder=decoder, multi_scale_discriminator=None, multi_period_discriminator=None, multi_scale_stft_discriminator=MultiScaleSTFTDiscriminator( n_filters=params.stft_discriminator_n_filters, n_ffts=params.n_ffts, hop_lengths=params.hop_lengths, win_lengths=params.win_lengths, ), ) return model def prepare_input( params: AttributeDict, batch: dict, device: torch.device, is_training: bool = True, ): """Parse batch data""" audio = batch["audio"].to(device, memory_format=torch.contiguous_format) features = batch["features"].to(device, memory_format=torch.contiguous_format) audio_lens = batch["audio_lens"].to(device) features_lens = batch["features_lens"].to(device) if is_training: audio_dims = audio.size(-1) start_idx = random.randint(0, max(0, audio_dims - params.sampling_rate)) audio = audio[:, start_idx : params.sampling_rate + start_idx] else: # NOTE(zengrui): a very coarse setup audio = audio[ :, params.sampling_rate : params.sampling_rate + params.sampling_rate ] if params.audio_normalization: mean = audio.mean(dim=-1, keepdim=True) std = audio.std(dim=-1, keepdim=True) audio = (audio - mean) / (std + 1e-7) return audio, audio_lens, features, features_lens def train_discriminator(weight, global_step, threshold=0, value=0.0): if global_step < threshold: weight = value return weight def train_one_epoch( params: AttributeDict, model: Union[nn.Module, DDP], optimizer_g: Optimizer, optimizer_d: Optimizer, scheduler_g: LRSchedulerType, scheduler_d: LRSchedulerType, train_dl: torch.utils.data.DataLoader, valid_dl: torch.utils.data.DataLoader, scaler: GradScaler, tb_writer: Optional[SummaryWriter] = None, world_size: int = 1, rank: int = 0, ) -> None: """Train the model for one epoch. The training loss from the mean of all frames is saved in `params.train_loss`. It runs the validation process every `params.valid_interval` batches. Args: params: It is returned by :func:`get_params`. model: The model to be trained. optimizer_g: The optimizer for generator. optimizer_d: The optimizer for discriminator. scheduler_g: The learning rate scheduler for generator, we call step() every epoch. scheduler_d: The learning rate scheduler for discriminator, we call step() every epoch. train_dl: Dataloader for the training dataset. valid_dl: Dataloader for the validation dataset. scaler: The scaler used for mix precision training. tb_writer: Writer to write log messages to tensorboard. world_size: Number of nodes in DDP training. If it is 1, DDP is disabled. rank: The rank of the node in DDP training. If no DDP is used, it should be set to 0. """ model.train() device = model.device if isinstance(model, DDP) else next(model.parameters()).device # used to summary the stats over iterations in one epoch tot_loss = MetricsTracker() saved_bad_model = False def save_bad_model(suffix: str = ""): save_checkpoint( filename=params.exp_dir / f"bad-model{suffix}-{rank}.pt", model=model, params=params, optimizer_g=optimizer_g, optimizer_d=optimizer_d, scheduler_g=scheduler_g, scheduler_d=scheduler_d, sampler=train_dl.sampler, scaler=scaler, rank=0, ) for batch_idx, batch in enumerate(train_dl): params.batch_idx_train += 1 batch_size = len(batch["audio"]) ( audio, audio_lens, _, _, ) = prepare_input(params, batch, device) loss_info = MetricsTracker() loss_info["samples"] = batch_size try: with autocast(enabled=params.use_fp16): d_weight = train_discriminator( params.lambda_adv, params.cur_epoch, threshold=params.discriminator_epoch_start, ) # forward discriminator ( disc_stft_real_adv_loss, disc_stft_fake_adv_loss, disc_period_real_adv_loss, disc_period_fake_adv_loss, disc_scale_real_adv_loss, disc_scale_fake_adv_loss, stats_d, ) = model( speech=audio, speech_lengths=audio_lens, return_sample=False, forward_generator=False, ) disc_loss = ( disc_stft_real_adv_loss + disc_stft_fake_adv_loss + disc_period_real_adv_loss + disc_period_fake_adv_loss + disc_scale_real_adv_loss + disc_scale_fake_adv_loss ) * d_weight for k, v in stats_d.items(): loss_info[k] = v * batch_size # update discriminator optimizer_d.zero_grad() scaler.scale(disc_loss).backward() scaler.step(optimizer_d) with autocast(enabled=params.use_fp16): g_weight = train_discriminator( params.lambda_adv, params.cur_epoch, threshold=params.discriminator_epoch_start, ) # forward generator ( commit_loss, gen_stft_adv_loss, gen_period_adv_loss, gen_scale_adv_loss, feature_stft_loss, feature_period_loss, feature_scale_loss, wav_reconstruction_loss, mel_reconstruction_loss, stats_g, ) = model( speech=audio, speech_lengths=audio_lens, forward_generator=True, return_sample=params.batch_idx_train % params.log_interval == 0, ) gen_adv_loss = ( gen_stft_adv_loss + gen_period_adv_loss + gen_scale_adv_loss ) * g_weight feature_loss = ( feature_stft_loss + feature_period_loss + feature_scale_loss ) reconstruction_loss = ( params.lambda_wav * wav_reconstruction_loss + params.lambda_rec * mel_reconstruction_loss ) gen_loss = ( gen_adv_loss + reconstruction_loss + params.lambda_feat * feature_loss + params.lambda_com * commit_loss ) loss_info["generator_loss"] = gen_loss for k, v in stats_g.items(): if "returned_sample" not in k: loss_info[k] = v * batch_size # update generator optimizer_g.zero_grad() scaler.scale(gen_loss).backward() scaler.step(optimizer_g) scaler.update() # summary stats tot_loss = tot_loss + loss_info except: # noqa save_bad_model() raise # step per iteration scheduler_g.step() scheduler_d.step() if params.print_diagnostics and batch_idx == 5: return if params.batch_idx_train % 100 == 0 and params.use_fp16: # If the grad scale was less than 1, try increasing it. The _growth_interval # of the grad scaler is configurable, but we can't configure it to have different # behavior depending on the current grad scale. cur_grad_scale = scaler._scale.item() if cur_grad_scale < 8.0 or ( cur_grad_scale < 32.0 and params.batch_idx_train % 400 == 0 ): scaler.update(cur_grad_scale * 2.0) if cur_grad_scale < 0.01: if not saved_bad_model: save_bad_model(suffix="-first-warning") saved_bad_model = True logging.warning(f"Grad scale is small: {cur_grad_scale}") if cur_grad_scale < 1.0e-05: save_bad_model() raise RuntimeError( f"grad_scale is too small, exiting: {cur_grad_scale}" ) if params.batch_idx_train % params.log_interval == 0: cur_lr_g = max(scheduler_g.get_last_lr()) cur_lr_d = max(scheduler_d.get_last_lr()) cur_grad_scale = scaler._scale.item() if params.use_fp16 else 1.0 logging.info( f"Epoch {params.cur_epoch}, batch {batch_idx}, " f"global_batch_idx: {params.batch_idx_train}, batch size: {batch_size}, " f"loss[{loss_info}], tot_loss[{tot_loss}], " f"cur_lr_g: {cur_lr_g:.2e}, cur_lr_d: {cur_lr_d:.2e}, " + (f"grad_scale: {scaler._scale.item()}" if params.use_fp16 else "") ) if tb_writer is not None: tb_writer.add_scalar( "train/learning_rate_g", cur_lr_g, params.batch_idx_train ) tb_writer.add_scalar( "train/learning_rate_d", cur_lr_d, params.batch_idx_train ) loss_info.write_summary( tb_writer, "train/current_", params.batch_idx_train ) tot_loss.write_summary(tb_writer, "train/tot_", params.batch_idx_train) if params.use_fp16: tb_writer.add_scalar( "train/grad_scale", cur_grad_scale, params.batch_idx_train ) if "returned_sample" in stats_g: # speech_hat_, speech_, mel_hat_, mel_ = stats_g["returned_sample"] speech_hat_, speech_, _, _ = stats_g["returned_sample"] speech_hat_i = speech_hat_[0] speech_i = speech_[0] if speech_hat_i.dim() > 1: speech_hat_i = speech_hat_i.squeeze(0) speech_i = speech_i.squeeze(0) tb_writer.add_audio( f"train/speech_hat_", speech_hat_i, params.batch_idx_train, params.sampling_rate, ) tb_writer.add_audio( f"train/speech_", speech_i, params.batch_idx_train, params.sampling_rate, ) # tb_writer.add_image( # "train/mel_hat_", # plot_feature(mel_hat_), # params.batch_idx_train, # dataformats="HWC", # ) # tb_writer.add_image( # "train/mel_", # plot_feature(mel_), # params.batch_idx_train, # dataformats="HWC", # ) if ( params.batch_idx_train % params.valid_interval == 0 and not params.print_diagnostics ): logging.info("Computing validation loss") valid_info, (speech_hat, speech) = compute_validation_loss( params=params, model=model, valid_dl=valid_dl, world_size=world_size, rank=rank, ) model.train() logging.info(f"Epoch {params.cur_epoch}, validation: {valid_info}") logging.info( f"Maximum memory allocated so far is {torch.cuda.max_memory_allocated()//1000000}MB" ) if tb_writer is not None and rank == 0: valid_info.write_summary( tb_writer, "train/valid_", params.batch_idx_train ) for index in range(params.num_samples): # 3 speech_hat_i = speech_hat[index] speech_i = speech[index] if speech_hat_i.dim() > 1: speech_hat_i = speech_hat_i.squeeze(0) speech_i = speech_i.squeeze(0) tb_writer.add_audio( f"train/valid_speech_hat_{index}", speech_hat_i, params.batch_idx_train, params.sampling_rate, ) tb_writer.add_audio( f"train/valid_speech_{index}", speech_i, params.batch_idx_train, params.sampling_rate, ) loss_value = tot_loss["generator_loss"] / tot_loss["samples"] params.train_loss = loss_value if params.train_loss < params.best_train_loss: params.best_train_epoch = params.cur_epoch params.best_train_loss = params.train_loss def compute_validation_loss( params: AttributeDict, model: Union[nn.Module, DDP], valid_dl: torch.utils.data.DataLoader, world_size: int = 1, rank: int = 0, ) -> Tuple[MetricsTracker, Tuple[np.ndarray, np.ndarray]]: """Run the validation process.""" model.eval() device = model.device if isinstance(model, DDP) else next(model.parameters()).device # used to summary the stats over iterations tot_loss = MetricsTracker() returned_sample = (None, None) with torch.no_grad(): for batch_idx, batch in enumerate(valid_dl): batch_size = len(batch["audio"]) ( audio, audio_lens, _, _, ) = prepare_input(params, batch, device, is_training=False) loss_info = MetricsTracker() loss_info["samples"] = batch_size d_weight = train_discriminator( params.lambda_adv, params.cur_epoch, threshold=params.discriminator_epoch_start, ) # forward discriminator ( disc_stft_real_adv_loss, disc_stft_fake_adv_loss, disc_period_real_adv_loss, disc_period_fake_adv_loss, disc_scale_real_adv_loss, disc_scale_fake_adv_loss, stats_d, ) = model( speech=audio, speech_lengths=audio_lens, return_sample=False, forward_generator=False, ) disc_loss = ( disc_stft_real_adv_loss + disc_stft_fake_adv_loss + disc_period_real_adv_loss + disc_period_fake_adv_loss + disc_scale_real_adv_loss + disc_scale_fake_adv_loss ) * d_weight assert disc_loss.requires_grad is False loss_info["discriminator_loss"] = disc_loss for k, v in stats_d.items(): loss_info[k] = v * batch_size g_weight = train_discriminator( params.lambda_adv, params.cur_epoch, threshold=params.discriminator_epoch_start, ) # forward generator ( commit_loss, gen_stft_adv_loss, gen_period_adv_loss, gen_scale_adv_loss, feature_stft_loss, feature_period_loss, feature_scale_loss, wav_reconstruction_loss, mel_reconstruction_loss, stats_g, ) = model( speech=audio, speech_lengths=audio_lens, forward_generator=True, return_sample=False, ) gen_adv_loss = ( gen_stft_adv_loss + gen_period_adv_loss + gen_scale_adv_loss ) * g_weight feature_loss = feature_stft_loss + feature_period_loss + feature_scale_loss reconstruction_loss = ( params.lambda_wav * wav_reconstruction_loss + params.lambda_rec * mel_reconstruction_loss ) gen_loss = ( gen_adv_loss + reconstruction_loss + params.lambda_feat * feature_loss + params.lambda_com * commit_loss ) assert gen_loss.requires_grad is False loss_info["generator_loss"] = gen_loss for k, v in stats_g.items(): if "returned_sample" not in k: loss_info[k] = v * batch_size # summary stats tot_loss = tot_loss + loss_info # infer for first batch: if batch_idx == 0 and rank == 0: inner_model = model.module if isinstance(model, DDP) else model _, audio_hat = inner_model.inference( x=audio, target_bw=params.target_bw ) returned_sample = (audio_hat, audio) if world_size > 1: tot_loss.reduce(device) loss_value = tot_loss["generator_loss"] / tot_loss["samples"] if loss_value < params.best_valid_loss: params.best_valid_epoch = params.cur_epoch params.best_valid_loss = loss_value return tot_loss, returned_sample def scan_pessimistic_batches_for_oom( model: Union[nn.Module, DDP], train_dl: torch.utils.data.DataLoader, optimizer_g: torch.optim.Optimizer, optimizer_d: torch.optim.Optimizer, params: AttributeDict, ): from lhotse.dataset import find_pessimistic_batches logging.info( "Sanity check -- see if any of the batches in epoch 1 would cause OOM." ) device = model.device if isinstance(model, DDP) else next(model.parameters()).device batches, crit_values = find_pessimistic_batches(train_dl.sampler) for criterion, cuts in batches.items(): batch = train_dl.dataset[cuts] ( audio, audio_lens, _, _, ) = prepare_input(params, batch, device) try: # for discriminator with autocast(enabled=params.use_fp16): ( disc_stft_real_adv_loss, disc_stft_fake_adv_loss, disc_period_real_adv_loss, disc_period_fake_adv_loss, disc_scale_real_adv_loss, disc_scale_fake_adv_loss, stats_d, ) = model( speech=audio, speech_lengths=audio_lens, return_sample=False, forward_generator=False, ) loss_d = ( disc_stft_real_adv_loss + disc_stft_fake_adv_loss + disc_period_real_adv_loss + disc_period_fake_adv_loss + disc_scale_real_adv_loss + disc_scale_fake_adv_loss ) * train_discriminator( params.lambda_adv, params.cur_epoch, threshold=params.discriminator_train_start, ) optimizer_d.zero_grad() loss_d.backward() # for generator with autocast(enabled=params.use_fp16): ( commit_loss, gen_stft_adv_loss, gen_period_adv_loss, gen_scale_adv_loss, feature_stft_loss, feature_period_loss, feature_scale_loss, wav_reconstruction_loss, mel_reconstruction_loss, stats_g, ) = model( speech=audio, speech_lengths=audio_lens, forward_generator=True, return_sample=False, ) loss_g = ( (gen_stft_adv_loss + gen_period_adv_loss + gen_scale_adv_loss) * train_discriminator( params.lambda_adv, 0, threshold=params.discriminator_epoch_start, ) + ( params.lambda_wav * wav_reconstruction_loss + params.lambda_rec * mel_reconstruction_loss ) + params.lambda_feat * (feature_stft_loss + feature_period_loss + feature_scale_loss) + params.lambda_com * commit_loss ) optimizer_g.zero_grad() loss_g.backward() except Exception as e: if "CUDA out of memory" in str(e): logging.error( "Your GPU ran out of memory with the current " "max_duration setting. We recommend decreasing " "max_duration and trying again.\n" f"Failing criterion: {criterion} " f"(={crit_values[criterion]}) ..." ) raise logging.info( f"Maximum memory allocated so far is {torch.cuda.max_memory_allocated()//1000000}MB" ) def run(rank, world_size, args): """ Args: rank: It is a value between 0 and `world_size-1`, which is passed automatically by `mp.spawn()` in :func:`main`. The node with rank 0 is responsible for saving checkpoint. world_size: Number of GPUs for DDP training. args: The return value of get_parser().parse_args() """ params = get_params() params.update(vars(args)) fix_random_seed(params.seed) if world_size > 1: setup_dist(rank, world_size, params.master_port) setup_logger(f"{params.exp_dir}/log/log-train") logging.info("Training started") if args.tensorboard and rank == 0: tb_writer = SummaryWriter(log_dir=f"{params.exp_dir}/tensorboard") else: tb_writer = None device = torch.device("cpu") if torch.cuda.is_available(): device = torch.device("cuda", rank) logging.info(f"Device: {device}") libritts = LibriTTSCodecDataModule(args) if params.full_libri: train_cuts = libritts.train_all_shuf_cuts() else: train_cuts = libritts.train_clean_100_cuts() logging.info(params) logging.info("About to create model") model = get_model(params) encoder = model.encoder decoder = model.decoder quantizer = model.quantizer multi_scale_discriminator = model.multi_scale_discriminator multi_period_discriminator = model.multi_period_discriminator multi_scale_stft_discriminator = model.multi_scale_stft_discriminator num_param_e = sum([p.numel() for p in encoder.parameters()]) logging.info(f"Number of parameters in encoder: {num_param_e}") num_param_d = sum([p.numel() for p in decoder.parameters()]) logging.info(f"Number of parameters in decoder: {num_param_d}") num_param_q = sum([p.numel() for p in quantizer.parameters()]) logging.info(f"Number of parameters in quantizer: {num_param_q}") num_param_ds = ( sum([p.numel() for p in multi_scale_discriminator.parameters()]) if multi_scale_discriminator is not None else 0 ) logging.info(f"Number of parameters in multi_scale_discriminator: {num_param_ds}") num_param_dp = ( sum([p.numel() for p in multi_period_discriminator.parameters()]) if multi_period_discriminator is not None else 0 ) logging.info(f"Number of parameters in multi_period_discriminator: {num_param_dp}") num_param_dstft = sum( [p.numel() for p in multi_scale_stft_discriminator.parameters()] ) logging.info( f"Number of parameters in multi_scale_stft_discriminator: {num_param_dstft}" ) logging.info( f"Total number of parameters: {num_param_e + num_param_d + num_param_q + num_param_ds + num_param_dp + num_param_dstft}" ) assert params.start_epoch > 0, params.start_epoch checkpoints = load_checkpoint_if_available(params=params, model=model) model.to(device) if world_size > 1: logging.info("Using DDP") model = nn.SyncBatchNorm.convert_sync_batchnorm(model) model = DDP( model, device_ids=[rank], find_unused_parameters=True, ) optimizer_g = torch.optim.AdamW( itertools.chain( encoder.parameters(), quantizer.parameters(), decoder.parameters(), ), lr=params.lr, betas=(0.5, 0.9), ) discriminator_params = [ multi_scale_stft_discriminator.parameters(), ] if multi_scale_discriminator is not None: discriminator_params.append(multi_scale_discriminator.parameters()) if multi_period_discriminator is not None: discriminator_params.append(multi_period_discriminator.parameters()) optimizer_d = torch.optim.AdamW( itertools.chain(*discriminator_params), lr=params.lr, betas=(0.5, 0.9), ) scheduler_g = WarmupCosineLrScheduler( optimizer=optimizer_g, max_iter=params.num_epochs * 1500, eta_ratio=0.1, warmup_iter=params.discriminator_epoch_start * 1500, warmup_ratio=1e-4, ) scheduler_d = WarmupCosineLrScheduler( optimizer=optimizer_d, max_iter=params.num_epochs * 1500, eta_ratio=0.1, warmup_iter=params.discriminator_epoch_start * 1500, warmup_ratio=1e-4, ) if checkpoints is not None: # load state_dict for optimizers if "optimizer_g" in checkpoints: logging.info("Loading optimizer_g state dict") optimizer_g.load_state_dict(checkpoints["optimizer_g"]) if "optimizer_d" in checkpoints: logging.info("Loading optimizer_d state dict") optimizer_d.load_state_dict(checkpoints["optimizer_d"]) # load state_dict for schedulers if "scheduler_g" in checkpoints: logging.info("Loading scheduler_g state dict") scheduler_g.load_state_dict(checkpoints["scheduler_g"]) if "scheduler_d" in checkpoints: logging.info("Loading scheduler_d state dict") scheduler_d.load_state_dict(checkpoints["scheduler_d"]) if params.print_diagnostics: opts = diagnostics.TensorDiagnosticOptions( 512 ) # allow 4 megabytes per sub-module diagnostic = diagnostics.attach_diagnostics(model, opts) if params.inf_check: register_inf_check_hooks(model) train_dl = libritts.train_dataloaders( train_cuts, world_size=world_size, rank=rank, ) valid_cuts = libritts.dev_clean_cuts() valid_dl = libritts.valid_dataloaders( valid_cuts, world_size=world_size, rank=rank, ) if not params.print_diagnostics: scan_pessimistic_batches_for_oom( model=model, train_dl=train_dl, optimizer_g=optimizer_g, optimizer_d=optimizer_d, params=params, ) scaler = GradScaler(enabled=params.use_fp16, init_scale=1.0) if checkpoints and "grad_scaler" in checkpoints: logging.info("Loading grad scaler state dict") scaler.load_state_dict(checkpoints["grad_scaler"]) for epoch in range(params.start_epoch, params.num_epochs + 1): logging.info(f"Start epoch {epoch}") fix_random_seed(params.seed + epoch - 1) train_dl.sampler.set_epoch(epoch - 1) params.cur_epoch = epoch if tb_writer is not None: tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train) train_one_epoch( params=params, model=model, optimizer_g=optimizer_g, optimizer_d=optimizer_d, scheduler_g=scheduler_g, scheduler_d=scheduler_d, train_dl=train_dl, valid_dl=valid_dl, scaler=scaler, tb_writer=tb_writer, world_size=world_size, rank=rank, ) if params.print_diagnostics: diagnostic.print_diagnostics() break if epoch % params.save_every_n == 0 or epoch == params.num_epochs: filename = params.exp_dir / f"epoch-{params.cur_epoch}.pt" save_checkpoint( filename=filename, params=params, model=model, optimizer_g=optimizer_g, optimizer_d=optimizer_d, scheduler_g=scheduler_g, scheduler_d=scheduler_d, sampler=train_dl.sampler, scaler=scaler, rank=rank, ) if rank == 0: if params.best_train_epoch == params.cur_epoch: best_train_filename = params.exp_dir / "best-train-loss.pt" copyfile(src=filename, dst=best_train_filename) if params.best_valid_epoch == params.cur_epoch: best_valid_filename = params.exp_dir / "best-valid-loss.pt" copyfile(src=filename, dst=best_valid_filename) logging.info("Done!") if world_size > 1: torch.distributed.barrier() cleanup_dist() def main(): parser = get_parser() LibriTTSCodecDataModule.add_arguments(parser) args = parser.parse_args() args.exp_dir = Path(args.exp_dir) world_size = args.world_size assert world_size >= 1 if world_size > 1: mp.spawn(run, args=(world_size, args), nprocs=world_size, join=True) else: run(rank=0, world_size=1, args=args) torch.set_num_threads(1) torch.set_num_interop_threads(1) if __name__ == "__main__": main()