#!/usr/bin/env python3 # -*- coding: utf-8 -*- import os import sys import time import traceback from random import randrange import numpy as np import torch # DISTRIBUTED from torch.nn.parallel import DistributedDataParallel as DDP_th from torch.utils.data import DataLoader from torch.utils.data.distributed import DistributedSampler from TTS.tts.datasets.preprocess import load_meta_data from TTS.tts.datasets.TTSDataset import MyDataset from TTS.tts.layers.losses import SpeedySpeechLoss from TTS.tts.utils.generic_utils import setup_model from TTS.tts.utils.io import save_best_model, save_checkpoint from TTS.tts.utils.measures import alignment_diagonal_score from TTS.tts.utils.speakers import parse_speakers from TTS.tts.utils.synthesis import synthesis from TTS.tts.utils.text.symbols import make_symbols, phonemes, symbols from TTS.tts.utils.visual import plot_alignment, plot_spectrogram from TTS.utils.arguments import init_training from TTS.utils.audio import AudioProcessor from TTS.utils.distribute import init_distributed, reduce_tensor from TTS.utils.generic_utils import KeepAverage, count_parameters, remove_experiment_folder, set_init_dict from TTS.utils.radam import RAdam from TTS.utils.training import NoamLR, setup_torch_training_env use_cuda, num_gpus = setup_torch_training_env(True, False) def setup_loader(ap, r, is_val=False, verbose=False): if is_val and not config.run_eval: loader = None else: dataset = MyDataset( r, config.text_cleaner, compute_linear_spec=False, meta_data=meta_data_eval if is_val else meta_data_train, ap=ap, tp=config.characters, add_blank=config["add_blank"], batch_group_size=0 if is_val else config.batch_group_size * config.batch_size, min_seq_len=config.min_seq_len, max_seq_len=config.max_seq_len, phoneme_cache_path=config.phoneme_cache_path, use_phonemes=config.use_phonemes, phoneme_language=config.phoneme_language, enable_eos_bos=config.enable_eos_bos_chars, use_noise_augment=not is_val, verbose=verbose, speaker_mapping=speaker_mapping if config.use_speaker_embedding and config.use_external_speaker_embedding_file else None, ) if config.use_phonemes and config.compute_input_seq_cache: # precompute phonemes to have a better estimate of sequence lengths. dataset.compute_input_seq(config.num_loader_workers) dataset.sort_items() sampler = DistributedSampler(dataset) if num_gpus > 1 else None loader = DataLoader( dataset, batch_size=config.eval_batch_size if is_val else config.batch_size, shuffle=False, collate_fn=dataset.collate_fn, drop_last=False, sampler=sampler, num_workers=config.num_val_loader_workers if is_val else config.num_loader_workers, pin_memory=False, ) return loader def format_data(data): # setup input data text_input = data[0] text_lengths = data[1] speaker_names = data[2] mel_input = data[4].permute(0, 2, 1) # B x D x T mel_lengths = data[5] item_idx = data[7] attn_mask = data[9] avg_text_length = torch.mean(text_lengths.float()) avg_spec_length = torch.mean(mel_lengths.float()) if config.use_speaker_embedding: if config.use_external_speaker_embedding_file: # return precomputed embedding vector speaker_c = data[8] else: # return speaker_id to be used by an embedding layer speaker_c = [speaker_mapping[speaker_name] for speaker_name in speaker_names] speaker_c = torch.LongTensor(speaker_c) else: speaker_c = None # compute durations from attention mask durations = torch.zeros(attn_mask.shape[0], attn_mask.shape[2]) for idx, am in enumerate(attn_mask): # compute raw durations c_idxs = am[:, : text_lengths[idx], : mel_lengths[idx]].max(1)[1] # c_idxs, counts = torch.unique_consecutive(c_idxs, return_counts=True) c_idxs, counts = torch.unique(c_idxs, return_counts=True) dur = torch.ones([text_lengths[idx]]).to(counts.dtype) dur[c_idxs] = counts # smooth the durations and set any 0 duration to 1 # by cutting off from the largest duration indeces. extra_frames = dur.sum() - mel_lengths[idx] largest_idxs = torch.argsort(-dur)[:extra_frames] dur[largest_idxs] -= 1 assert ( dur.sum() == mel_lengths[idx] ), f" [!] total duration {dur.sum()} vs spectrogram length {mel_lengths[idx]}" durations[idx, : text_lengths[idx]] = dur # dispatch data to GPU if use_cuda: text_input = text_input.cuda(non_blocking=True) text_lengths = text_lengths.cuda(non_blocking=True) mel_input = mel_input.cuda(non_blocking=True) mel_lengths = mel_lengths.cuda(non_blocking=True) if speaker_c is not None: speaker_c = speaker_c.cuda(non_blocking=True) attn_mask = attn_mask.cuda(non_blocking=True) durations = durations.cuda(non_blocking=True) return ( text_input, text_lengths, mel_input, mel_lengths, speaker_c, avg_text_length, avg_spec_length, attn_mask, durations, item_idx, ) def train(data_loader, model, criterion, optimizer, scheduler, ap, global_step, epoch): model.train() epoch_time = 0 keep_avg = KeepAverage() if use_cuda: batch_n_iter = int(len(data_loader.dataset) / (config.batch_size * num_gpus)) else: batch_n_iter = int(len(data_loader.dataset) / config.batch_size) end_time = time.time() c_logger.print_train_start() scaler = torch.cuda.amp.GradScaler() if config.mixed_precision else None for num_iter, data in enumerate(data_loader): start_time = time.time() # format data ( text_input, text_lengths, mel_targets, mel_lengths, speaker_c, avg_text_length, avg_spec_length, _, dur_target, _, ) = format_data(data) loader_time = time.time() - end_time global_step += 1 optimizer.zero_grad() # forward pass model with torch.cuda.amp.autocast(enabled=config.mixed_precision): decoder_output, dur_output, alignments = model.forward( text_input, text_lengths, mel_lengths, dur_target, g=speaker_c ) # compute loss loss_dict = criterion( decoder_output, mel_targets, mel_lengths, dur_output, torch.log(1 + dur_target), text_lengths ) # backward pass with loss scaling if config.mixed_precision: scaler.scale(loss_dict["loss"]).backward() scaler.unscale_(optimizer) grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), config.grad_clip) scaler.step(optimizer) scaler.update() else: loss_dict["loss"].backward() grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), config.grad_clip) optimizer.step() # setup lr if config.noam_schedule: scheduler.step() # current_lr current_lr = optimizer.param_groups[0]["lr"] # compute alignment error (the lower the better ) align_error = 1 - alignment_diagonal_score(alignments, binary=True) loss_dict["align_error"] = align_error step_time = time.time() - start_time epoch_time += step_time # aggregate losses from processes if num_gpus > 1: loss_dict["loss_l1"] = reduce_tensor(loss_dict["loss_l1"].data, num_gpus) loss_dict["loss_ssim"] = reduce_tensor(loss_dict["loss_ssim"].data, num_gpus) loss_dict["loss_dur"] = reduce_tensor(loss_dict["loss_dur"].data, num_gpus) loss_dict["loss"] = reduce_tensor(loss_dict["loss"].data, num_gpus) # detach loss values loss_dict_new = dict() for key, value in loss_dict.items(): if isinstance(value, (int, float)): loss_dict_new[key] = value else: loss_dict_new[key] = value.item() loss_dict = loss_dict_new # update avg stats update_train_values = dict() for key, value in loss_dict.items(): update_train_values["avg_" + key] = value update_train_values["avg_loader_time"] = loader_time update_train_values["avg_step_time"] = step_time keep_avg.update_values(update_train_values) # print training progress if global_step % config.print_step == 0: log_dict = { "avg_spec_length": [avg_spec_length, 1], # value, precision "avg_text_length": [avg_text_length, 1], "step_time": [step_time, 4], "loader_time": [loader_time, 2], "current_lr": current_lr, } c_logger.print_train_step(batch_n_iter, num_iter, global_step, log_dict, loss_dict, keep_avg.avg_values) if args.rank == 0: # Plot Training Iter Stats # reduce TB load if global_step % config.tb_plot_step == 0: iter_stats = {"lr": current_lr, "grad_norm": grad_norm, "step_time": step_time} iter_stats.update(loss_dict) tb_logger.tb_train_iter_stats(global_step, iter_stats) if global_step % config.save_step == 0: if config.checkpoint: # save model save_checkpoint( model, optimizer, global_step, epoch, 1, OUT_PATH, model_characters, model_loss=loss_dict["loss"], ) # wait all kernels to be completed torch.cuda.synchronize() # Diagnostic visualizations idx = np.random.randint(mel_targets.shape[0]) pred_spec = decoder_output[idx].detach().data.cpu().numpy().T gt_spec = mel_targets[idx].data.cpu().numpy().T align_img = alignments[idx].data.cpu() figures = { "prediction": plot_spectrogram(pred_spec, ap), "ground_truth": plot_spectrogram(gt_spec, ap), "alignment": plot_alignment(align_img), } tb_logger.tb_train_figures(global_step, figures) # Sample audio train_audio = ap.inv_melspectrogram(pred_spec.T) tb_logger.tb_train_audios(global_step, {"TrainAudio": train_audio}, config.audio["sample_rate"]) end_time = time.time() # print epoch stats c_logger.print_train_epoch_end(global_step, epoch, epoch_time, keep_avg) # Plot Epoch Stats if args.rank == 0: epoch_stats = {"epoch_time": epoch_time} epoch_stats.update(keep_avg.avg_values) tb_logger.tb_train_epoch_stats(global_step, epoch_stats) if config.tb_model_param_stats: tb_logger.tb_model_weights(model, global_step) return keep_avg.avg_values, global_step @torch.no_grad() def evaluate(data_loader, model, criterion, ap, global_step, epoch): model.eval() epoch_time = 0 keep_avg = KeepAverage() c_logger.print_eval_start() if data_loader is not None: for num_iter, data in enumerate(data_loader): start_time = time.time() # format data text_input, text_lengths, mel_targets, mel_lengths, speaker_c, _, _, _, dur_target, _ = format_data(data) # forward pass model with torch.cuda.amp.autocast(enabled=config.mixed_precision): decoder_output, dur_output, alignments = model.forward( text_input, text_lengths, mel_lengths, dur_target, g=speaker_c ) # compute loss loss_dict = criterion( decoder_output, mel_targets, mel_lengths, dur_output, torch.log(1 + dur_target), text_lengths ) # step time step_time = time.time() - start_time epoch_time += step_time # compute alignment score align_error = 1 - alignment_diagonal_score(alignments, binary=True) loss_dict["align_error"] = align_error # aggregate losses from processes if num_gpus > 1: loss_dict["loss_l1"] = reduce_tensor(loss_dict["loss_l1"].data, num_gpus) loss_dict["loss_ssim"] = reduce_tensor(loss_dict["loss_ssim"].data, num_gpus) loss_dict["loss_dur"] = reduce_tensor(loss_dict["loss_dur"].data, num_gpus) loss_dict["loss"] = reduce_tensor(loss_dict["loss"].data, num_gpus) # detach loss values loss_dict_new = dict() for key, value in loss_dict.items(): if isinstance(value, (int, float)): loss_dict_new[key] = value else: loss_dict_new[key] = value.item() loss_dict = loss_dict_new # update avg stats update_train_values = dict() for key, value in loss_dict.items(): update_train_values["avg_" + key] = value keep_avg.update_values(update_train_values) if config.print_eval: c_logger.print_eval_step(num_iter, loss_dict, keep_avg.avg_values) if args.rank == 0: # Diagnostic visualizations idx = np.random.randint(mel_targets.shape[0]) pred_spec = decoder_output[idx].detach().data.cpu().numpy().T gt_spec = mel_targets[idx].data.cpu().numpy().T align_img = alignments[idx].data.cpu() eval_figures = { "prediction": plot_spectrogram(pred_spec, ap, output_fig=False), "ground_truth": plot_spectrogram(gt_spec, ap, output_fig=False), "alignment": plot_alignment(align_img, output_fig=False), } # Sample audio eval_audio = ap.inv_melspectrogram(pred_spec.T) tb_logger.tb_eval_audios(global_step, {"ValAudio": eval_audio}, config.audio["sample_rate"]) # Plot Validation Stats tb_logger.tb_eval_stats(global_step, keep_avg.avg_values) tb_logger.tb_eval_figures(global_step, eval_figures) if args.rank == 0 and epoch >= config.test_delay_epochs: if config.test_sentences_file: with open(config.test_sentences_file, "r") as f: test_sentences = [s.strip() for s in f.readlines()] else: test_sentences = [ "It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.", "Be a voice, not an echo.", "I'm sorry Dave. I'm afraid I can't do that.", "This cake is great. It's so delicious and moist.", "Prior to November 22, 1963.", ] # test sentences test_audios = {} test_figures = {} print(" | > Synthesizing test sentences") if config.use_speaker_embedding: if config.use_external_speaker_embedding_file: speaker_embedding = speaker_mapping[list(speaker_mapping.keys())[randrange(len(speaker_mapping) - 1)]][ "embedding" ] speaker_id = None else: speaker_id = 0 speaker_embedding = None else: speaker_id = None speaker_embedding = None for idx, test_sentence in enumerate(test_sentences): try: wav, alignment, _, postnet_output, _, _ = synthesis( model, test_sentence, config, use_cuda, ap, speaker_id=speaker_id, speaker_embedding=speaker_embedding, style_wav=None, truncated=False, enable_eos_bos_chars=config.enable_eos_bos_chars, # pylint: disable=unused-argument use_griffin_lim=True, do_trim_silence=False, ) file_path = os.path.join(AUDIO_PATH, str(global_step)) os.makedirs(file_path, exist_ok=True) file_path = os.path.join(file_path, "TestSentence_{}.wav".format(idx)) ap.save_wav(wav, file_path) test_audios["{}-audio".format(idx)] = wav test_figures["{}-prediction".format(idx)] = plot_spectrogram(postnet_output, ap) test_figures["{}-alignment".format(idx)] = plot_alignment(alignment) except: # pylint: disable=bare-except print(" !! Error creating Test Sentence -", idx) traceback.print_exc() tb_logger.tb_test_audios(global_step, test_audios, config.audio["sample_rate"]) tb_logger.tb_test_figures(global_step, test_figures) return keep_avg.avg_values # FIXME: move args definition/parsing inside of main? def main(args): # pylint: disable=redefined-outer-name # pylint: disable=global-variable-undefined global meta_data_train, meta_data_eval, symbols, phonemes, model_characters, speaker_mapping # Audio processor ap = AudioProcessor(**config.audio.to_dict()) if config.characters is not None: symbols, phonemes = make_symbols(**config.characters.to_dict()) # DISTRUBUTED if num_gpus > 1: init_distributed(args.rank, num_gpus, args.group_id, config.distributed["backend"], config.distributed["url"]) # set model characters model_characters = phonemes if config.use_phonemes else symbols num_chars = len(model_characters) # load data instances meta_data_train, meta_data_eval = load_meta_data(config.datasets, eval_split=True) # set the portion of the data used for training if set in config.json if config.has("train_portion"): meta_data_train = meta_data_train[: int(len(meta_data_train) * config.train_portion)] if config.has("eval_portion"): meta_data_eval = meta_data_eval[: int(len(meta_data_eval) * config.eval_portion)] # parse speakers num_speakers, speaker_embedding_dim, speaker_mapping = parse_speakers(config, args, meta_data_train, OUT_PATH) # setup model model = setup_model(num_chars, num_speakers, config, speaker_embedding_dim=speaker_embedding_dim) optimizer = RAdam(model.parameters(), lr=config.lr, weight_decay=0, betas=(0.9, 0.98), eps=1e-9) criterion = SpeedySpeechLoss(config) if args.restore_path: print(f" > Restoring from {os.path.basename(args.restore_path)} ...") checkpoint = torch.load(args.restore_path, map_location="cpu") try: # TODO: fix optimizer init, model.cuda() needs to be called before # optimizer restore optimizer.load_state_dict(checkpoint["optimizer"]) if config.reinit_layers: raise RuntimeError model.load_state_dict(checkpoint["model"]) except: # pylint: disable=bare-except print(" > Partial model initialization.") model_dict = model.state_dict() model_dict = set_init_dict(model_dict, checkpoint["model"], config) model.load_state_dict(model_dict) del model_dict for group in optimizer.param_groups: group["initial_lr"] = config.lr print(" > Model restored from step %d" % checkpoint["step"], flush=True) args.restore_step = checkpoint["step"] else: args.restore_step = 0 if use_cuda: model.cuda() criterion.cuda() # DISTRUBUTED if num_gpus > 1: model = DDP_th(model, device_ids=[args.rank]) if config.noam_schedule: scheduler = NoamLR(optimizer, warmup_steps=config.warmup_steps, last_epoch=args.restore_step - 1) else: scheduler = None num_params = count_parameters(model) print("\n > Model has {} parameters".format(num_params), flush=True) if args.restore_step == 0 or not args.best_path: best_loss = float("inf") print(" > Starting with inf best loss.") else: print(" > Restoring best loss from " f"{os.path.basename(args.best_path)} ...") best_loss = torch.load(args.best_path, map_location="cpu")["model_loss"] print(f" > Starting with loaded last best loss {best_loss}.") keep_all_best = config.keep_all_best keep_after = config.keep_after # void if keep_all_best False # define dataloaders train_loader = setup_loader(ap, 1, is_val=False, verbose=True) eval_loader = setup_loader(ap, 1, is_val=True, verbose=True) global_step = args.restore_step for epoch in range(0, config.epochs): c_logger.print_epoch_start(epoch, config.epochs) train_avg_loss_dict, global_step = train( train_loader, model, criterion, optimizer, scheduler, ap, global_step, epoch ) eval_avg_loss_dict = evaluate(eval_loader, model, criterion, ap, global_step, epoch) c_logger.print_epoch_end(epoch, eval_avg_loss_dict) target_loss = train_avg_loss_dict["avg_loss"] if config.run_eval: target_loss = eval_avg_loss_dict["avg_loss"] best_loss = save_best_model( target_loss, best_loss, model, optimizer, global_step, epoch, config.r, OUT_PATH, model_characters, keep_all_best=keep_all_best, keep_after=keep_after, ) if __name__ == "__main__": args, config, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = init_training(sys.argv) try: main(args) except KeyboardInterrupt: remove_experiment_folder(OUT_PATH) try: sys.exit(0) except SystemExit: os._exit(0) # pylint: disable=protected-access except Exception: # pylint: disable=broad-except remove_experiment_folder(OUT_PATH) traceback.print_exc() sys.exit(1)