import time import os import random import numpy as np import torch import torch.utils.data from tqdm import tqdm from module import commons from module.aligner import beta_binomial_prior_distribution from module.mel_processing import spectrogram_torch from text import cleaned_text_to_sequence from text.symbols import v from utils import load_wav_to_torch, load_filepaths_and_text import torch.nn.functional as F """Multi speaker version""" class TextAudioSpeakerLoader(torch.utils.data.Dataset): """ 1) loads audio, speaker_id, text pairs 2) normalizes text and converts them to sequences of integers 3) computes spectrograms from audio files. """ def __init__(self, audiopaths_sid_text, hparams, get_path=False, meta=None, val=False, phoneme_path='dump/phoneme.npy'): self.audiopaths_sid_text = load_filepaths_and_text(audiopaths_sid_text) self.max_wav_value = hparams.max_wav_value self.max_dur = hparams.max_dur self.sampling_rate = hparams.sampling_rate self.filter_length = hparams.filter_length self.hop_length = hparams.hop_length self.win_length = hparams.win_length self.sampling_rate = hparams.sampling_rate self.val = val self.get_path = get_path self.meta = meta self.phoneme_data = np.load(phoneme_path, allow_pickle=True).item() random.seed(1234) random.shuffle(self.audiopaths_sid_text) self._filter() def _filter(self): """ Filter text & store spec lengths """ # Store spectrogram lengths for Bucketing # wav_length ~= file_size / (wav_channels * Bytes per dim) = file_size / (1 * 2) # spec_length = wav_length // hop_length print('only chinese and skip other languages!!!!!!!!') print('only chinese and skip other languages!!!!!!!!') print('only chinese and skip other languages!!!!!!!!') print('only chinese and skip other languages!!!!!!!!') print("phoneme_data_len:", len(self.phoneme_data.keys())) print("wav_data_len:", len(self.audiopaths_sid_text)) audiopaths_sid_text_new = [] lengths = [] skipped = 0 for item in tqdm(self.audiopaths_sid_text): audiopath = item[0] if 'zh' not in audiopath: skipped += 1 continue try: phoneme = self.phoneme_data[audiopath] phoneme = phoneme.split(' ') phoneme_ids = cleaned_text_to_sequence(phoneme) except Exception: skipped += 1 continue bert_path = audiopath.replace('.wav', '.bert.pt').replace('.mp3', '.bert.pt') if not os.path.exists(bert_path): skipped += 1 continue sslpath = audiopath.replace('.wav', '.ssl.pt').replace('.mp3', '.ssl.pt') assert audiopath.endswith('.wav') dur = os.path.getsize(audiopath) / self.sampling_rate / 2 if os.path.exists(audiopath) and os.path.exists(sslpath) and ( 1 < dur < self.max_dur or self.val): audiopaths_sid_text_new.append([audiopath, sslpath, bert_path, phoneme_ids]) lengths.append(os.path.getsize(audiopath) // (2 * self.hop_length)) else: skipped += 1 print("skipped: ", skipped, ", total: ", len(self.audiopaths_sid_text)) self.audiopaths_sid_text = audiopaths_sid_text_new self.lengths = lengths def get_audio_text_speaker_pair(self, audiopath_sid_text): # separate filename, speaker_id and text audiopath, sslpath, bert_path, phoneme_ids = audiopath_sid_text # phoneme_ids = commons.intersperse(phoneme_ids, 0) res = [0] * (len(phoneme_ids) * 3) res[1::3] = phoneme_ids phoneme_ids = res text = torch.LongTensor(phoneme_ids) spec, wav = self.get_audio(audiopath) ssl = torch.load(sslpath) ssl = F.interpolate(ssl, size=spec.shape[-1], mode="nearest") bert_feature = torch.load(bert_path) bert_feature = F.interpolate(bert_feature.unsqueeze(0), scale_factor=3, mode='nearest') assert bert_feature.shape[-1] == len(phoneme_ids) spk_emb = np.load(audiopath.replace('.wav', '.spk.npy').replace('.mp3', '.spk.npy')) spk_emb = torch.FloatTensor(spk_emb) beta_binomial_scaling_factor = 1.0 # Calculate attention prior attn_prior = beta_binomial_prior_distribution( spec.shape[-1], text.shape[-1], beta_binomial_scaling_factor, ) return (ssl, spec, wav, text, bert_feature, spk_emb, attn_prior) def get_audio(self, filename): audio, sampling_rate = load_wav_to_torch(filename) if sampling_rate != self.sampling_rate: raise ValueError("{} SR doesn't match target {} SR".format( sampling_rate, self.sampling_rate)) audio_norm = audio audio_norm = audio_norm.unsqueeze(0) spec = spectrogram_torch(audio_norm, self.filter_length, self.sampling_rate, self.hop_length, self.win_length, center=False) spec = torch.squeeze(spec, 0) return spec, audio_norm def get_sid(self, sid): sid = torch.LongTensor([int(sid)]) return sid def __getitem__(self, index): return self.get_audio_text_speaker_pair(self.audiopaths_sid_text[index]) def __len__(self): return len(self.audiopaths_sid_text) class TextAudioSpeakerCollate(): """ Zero-pads model inputs and targets """ def pad_3D(self, inputs, B, T, L): inputs_padded = np.zeros((B, T, L), dtype=np.float32) for i, input_ in enumerate(inputs): shape0 = np.shape(input_)[0] shape1 = np.shape(input_)[1] if shape0 > T: print("pad error!!!\nerror!!!\n") # print(inputs.shape, inputs_padded.shape) shape0 = T if shape1 > L: print("pad error!!!\nerror!!!\n") # print(inputs.shape, inputs_padded.shape) shape1 = L inputs_padded[i, :shape0, :shape1] = input_[:shape0, :shape1] return torch.FloatTensor(inputs_padded) def __init__(self, return_ids=False): self.return_ids = return_ids def __call__(self, batch): """Collate's training batch from normalized text, audio and speaker identities PARAMS ------ batch: [text_normalized, spec_normalized, wav_normalized, sid] """ batch = [x for x in batch if x is not None] # Right zero-pad all one-hot text sequences to max input length _, ids_sorted_decreasing = torch.sort( torch.LongTensor([x[1].size(1) for x in batch]), dim=0, descending=True) max_spec_len = max([x[1].size(1) for x in batch]) max_wav_len = max([x[2].size(1) for x in batch]) max_text_len = max([x[3].size(0) for x in batch]) spec_lengths = torch.LongTensor(len(batch)) wav_lengths = torch.LongTensor(len(batch)) text_lengths = torch.LongTensor(len(batch)) spec_padded = torch.FloatTensor(len(batch), batch[0][1].size(0), max_spec_len) wav_padded = torch.FloatTensor(len(batch), 1, max_wav_len) ssl_padded = torch.FloatTensor(len(batch), 768, max_spec_len) text_padded = torch.LongTensor(len(batch), max_text_len) bert_padded = torch.FloatTensor(len(batch), 1024, max_text_len) spk_emb_padded = torch.FloatTensor(len(batch), 256) spec_padded.zero_() wav_padded.zero_() ssl_padded.zero_() text_padded.zero_() bert_padded.zero_() spk_emb_padded.zero_() attn_priors = [] for i in range(len(ids_sorted_decreasing)): row = batch[ids_sorted_decreasing[i]] ssl = row[0] ssl_padded[i, :, :ssl.size(2)] = ssl[0, :, :] spec = row[1] spec_padded[i, :, :spec.size(1)] = spec spec_lengths[i] = spec.size(1) wav = row[2] wav_padded[i, :, :wav.size(1)] = wav wav_lengths[i] = wav.size(1) text = row[3] text_padded[i, :text.size(0)] = text text_lengths[i] = text.size(0) bert = row[4] bert_padded[i, :, :bert.size(-1)] = bert spk_emb = row[5] spk_emb_padded[i] = spk_emb attn_priors.append(row[6]) attn_priors = self.pad_3D(attn_priors, len(batch), max_text_len, max_spec_len) return ssl_padded, spec_padded, spec_lengths, wav_padded, wav_lengths, text_padded, text_lengths, bert_padded, spk_emb_padded, attn_priors class DistributedBucketSampler(torch.utils.data.distributed.DistributedSampler): """ Maintain similar input lengths in a batch. Length groups are specified by boundaries. Ex) boundaries = [b1, b2, b3] -> any batch is included either {x | b1 < length(x) <=b2} or {x | b2 < length(x) <= b3}. It removes samples which are not included in the boundaries. Ex) boundaries = [b1, b2, b3] -> any x s.t. length(x) <= b1 or length(x) > b3 are discarded. """ def __init__(self, dataset, batch_size, boundaries, num_replicas=None, rank=None, shuffle=True): super().__init__(dataset, num_replicas=num_replicas, rank=rank, shuffle=shuffle) self.lengths = dataset.lengths self.batch_size = batch_size self.boundaries = boundaries self.buckets, self.num_samples_per_bucket = self._create_buckets() self.total_size = sum(self.num_samples_per_bucket) self.num_samples = self.total_size // self.num_replicas def _create_buckets(self): buckets = [[] for _ in range(len(self.boundaries) - 1)] for i in range(len(self.lengths)): length = self.lengths[i] idx_bucket = self._bisect(length) if idx_bucket != -1: buckets[idx_bucket].append(i) for i in range(len(buckets) - 1, 0, -1): if len(buckets[i]) == 0: buckets.pop(i) self.boundaries.pop(i + 1) num_samples_per_bucket = [] for i in range(len(buckets)): len_bucket = len(buckets[i]) total_batch_size = self.num_replicas * self.batch_size rem = (total_batch_size - (len_bucket % total_batch_size)) % total_batch_size num_samples_per_bucket.append(len_bucket + rem) return buckets, num_samples_per_bucket def __iter__(self): # deterministically shuffle based on epoch g = torch.Generator() g.manual_seed(self.epoch) indices = [] if self.shuffle: for bucket in self.buckets: indices.append(torch.randperm(len(bucket), generator=g).tolist()) else: for bucket in self.buckets: indices.append(list(range(len(bucket)))) batches = [] for i in range(len(self.buckets)): bucket = self.buckets[i] len_bucket = len(bucket) ids_bucket = indices[i] num_samples_bucket = self.num_samples_per_bucket[i] # add extra samples to make it evenly divisible rem = num_samples_bucket - len_bucket ids_bucket = ids_bucket + ids_bucket * (rem // len_bucket) + ids_bucket[:(rem % len_bucket)] # subsample ids_bucket = ids_bucket[self.rank::self.num_replicas] # batching for j in range(len(ids_bucket) // self.batch_size): batch = [bucket[idx] for idx in ids_bucket[j * self.batch_size:(j + 1) * self.batch_size]] batches.append(batch) if self.shuffle: batch_ids = torch.randperm(len(batches), generator=g).tolist() batches = [batches[i] for i in batch_ids] self.batches = batches assert len(self.batches) * self.batch_size == self.num_samples return iter(self.batches) def _bisect(self, x, lo=0, hi=None): if hi is None: hi = len(self.boundaries) - 1 if hi > lo: mid = (hi + lo) // 2 if self.boundaries[mid] < x and x <= self.boundaries[mid + 1]: return mid elif x <= self.boundaries[mid]: return self._bisect(x, lo, mid) else: return self._bisect(x, mid + 1, hi) else: return -1 def __len__(self): return self.num_samples // self.batch_size