Spaces:
Running
on
Zero
Running
on
Zero
File size: 6,009 Bytes
c968fc3 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 |
# Copyright (c) 2023 Amphion.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import torch
import random
import numpy as np
from torch.nn import functional as F
from torch.nn.utils.rnn import pad_sequence
from utils.data_utils import *
from models.vocoders.vocoder_dataset import VocoderDataset
class DiffusionVocoderDataset(VocoderDataset):
def __init__(self, cfg, dataset, is_valid=False):
"""
Args:
cfg: config
dataset: dataset name
is_valid: whether to use train or valid dataset
"""
super().__init__(cfg, dataset, is_valid)
eval_index = random.randint(0, len(self.metadata) - 1)
eval_utt_info = self.metadata[eval_index]
eval_utt = "{}_{}".format(eval_utt_info["Dataset"], eval_utt_info["Uid"])
self.eval_audio = np.load(self.utt2audio_path[eval_utt])
if cfg.preprocess.use_mel:
self.eval_mel = np.load(self.utt2mel_path[eval_utt])
if cfg.preprocess.use_frame_pitch:
self.eval_pitch = np.load(self.utt2frame_pitch_path[eval_utt])
def __getitem__(self, index):
utt_info = self.metadata[index]
dataset = utt_info["Dataset"]
uid = utt_info["Uid"]
utt = "{}_{}".format(dataset, uid)
single_feature = dict()
if self.cfg.preprocess.use_mel:
mel = np.load(self.utt2mel_path[utt])
assert mel.shape[0] == self.cfg.preprocess.n_mel
if "target_len" not in single_feature.keys():
single_feature["target_len"] = mel.shape[1]
if single_feature["target_len"] <= self.cfg.preprocess.cut_mel_frame:
mel = np.pad(
mel,
((0, 0), (0, self.cfg.preprocess.cut_mel_frame - mel.shape[-1])),
mode="constant",
)
else:
if "start" not in single_feature.keys():
start = random.randint(
0, mel.shape[-1] - self.cfg.preprocess.cut_mel_frame
)
end = start + self.cfg.preprocess.cut_mel_frame
single_feature["start"] = start
single_feature["end"] = end
mel = mel[:, single_feature["start"] : single_feature["end"]]
single_feature["mel"] = mel
if self.cfg.preprocess.use_frame_pitch:
frame_pitch = np.load(self.utt2frame_pitch_path[utt])
if "target_len" not in single_feature.keys():
single_feature["target_len"] = len(frame_pitch)
aligned_frame_pitch = align_length(
frame_pitch, single_feature["target_len"]
)
if single_feature["target_len"] <= self.cfg.preprocess.cut_mel_frame:
aligned_frame_pitch = np.pad(
aligned_frame_pitch,
(
(
0,
self.cfg.preprocess.cut_mel_frame
* self.cfg.preprocess.hop_size
- audio.shape[-1],
)
),
mode="constant",
)
else:
if "start" not in single_feature.keys():
start = random.randint(
0,
aligned_frame_pitch.shape[-1]
- self.cfg.preprocess.cut_mel_frame,
)
end = start + self.cfg.preprocess.cut_mel_frame
single_feature["start"] = start
single_feature["end"] = end
aligned_frame_pitch = aligned_frame_pitch[
single_feature["start"] : single_feature["end"]
]
single_feature["frame_pitch"] = aligned_frame_pitch
if self.cfg.preprocess.use_audio:
audio = np.load(self.utt2audio_path[utt])
assert "target_len" in single_feature.keys()
if (
audio.shape[-1]
<= self.cfg.preprocess.cut_mel_frame * self.cfg.preprocess.hop_size
):
audio = np.pad(
audio,
(
(
0,
self.cfg.preprocess.cut_mel_frame
* self.cfg.preprocess.hop_size
- audio.shape[-1],
)
),
mode="constant",
)
else:
if "start" not in single_feature.keys():
audio = audio[
0 : self.cfg.preprocess.cut_mel_frame
* self.cfg.preprocess.hop_size
]
else:
audio = audio[
single_feature["start"]
* self.cfg.preprocess.hop_size : single_feature["end"]
* self.cfg.preprocess.hop_size,
]
single_feature["audio"] = audio
return single_feature
class DiffusionVocoderCollator(object):
"""Zero-pads model inputs and targets based on number of frames per step"""
def __init__(self, cfg):
self.cfg = cfg
def __call__(self, batch):
packed_batch_features = dict()
# mel: [b, n_mels, frame]
# frame_pitch: [b, frame]
# audios: [b, frame * hop_size]
for key in batch[0].keys():
if key in ["target_len", "start", "end"]:
continue
else:
values = [torch.from_numpy(b[key]) for b in batch]
packed_batch_features[key] = pad_sequence(
values, batch_first=True, padding_value=0
)
return packed_batch_features
|