File size: 5,403 Bytes
4ee33aa
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import random

import torch
import torch.nn.functional as F
import torch.utils.data
from torch import LongTensor
from tqdm import tqdm
import torchaudio
from pypinyin import Style, lazy_pinyin
import math
from ttts.gpt.voice_tokenizer import VoiceBpeTokenizer
from ttts.utils.infer_utils import load_model
import json
import os

def padding_to_8(x):
    l = x.shape[-1]
    l = (math.floor(l / 8) + 1) * 8
    x = torch.nn.functional.pad(x, (0, l-x.shape[-1]))
    return x
def read_jsonl(path):
    with open(path, 'r') as f:
        json_str = f.read()
    data_list = []
    for line in json_str.splitlines():
        data = json.loads(line)
        data_list.append(data)
    return data_list
def write_jsonl(path, all_paths):
    with open(path,'w', encoding='utf-8') as file:
        for item in all_paths:
            json.dump(item, file, ensure_ascii=False)
            file.write('\n')

def padding_to_8(x):
    l = x.shape[-1]
    l = (math.floor(l / 8) + 1) * 8
    x = torch.nn.functional.pad(x, (0, l-x.shape[-1]))
    return x
class DiffusionDataset(torch.utils.data.Dataset):
    def __init__(self, opt):
        self.jsonl_path = opt['dataset']['path']
        self.audiopaths_and_text = read_jsonl(self.jsonl_path)
        self.tok = VoiceBpeTokenizer('ttts/gpt/gpt_tts_tokenizer.json')
    def __getitem__(self, index):
        # Fetch text and add start/stop tokens.
        audiopath_and_text = self.audiopaths_and_text[index]
        audiopath, text = audiopath_and_text['path'], audiopath_and_text['text']
        text = ' '.join(lazy_pinyin(text, style=Style.TONE3, neutral_tone_with_five=True))
        text = self.tok.encode(text)
        text_tokens = LongTensor(text)
        try:
            mel_path = audiopath + '.mel.pth'
            mel_raw = torch.load(mel_path)[0]

            quant_path = audiopath + '.melvq.pth'
            mel_codes = LongTensor(torch.load(quant_path)[0])
        except:
            return None

        # Define the number of frames for the random crop (adjust as needed)
        crop_frames = random.randint(int(mel_raw.shape[1] // 4), int(mel_raw.shape[1] // 4 * 3))

        # Ensure the crop doesn't exceed the length of the original audio
        max_start_frame = mel_raw.shape[1] - crop_frames
        start_frame = random.randint(0, max_start_frame)

        # Perform the random crop
        mel_refer = mel_raw[:, start_frame: start_frame + crop_frames]
        mel_refer = padding_to_8(mel_refer)
        # split = random.randint(int(mel_raw.shape[1]//3), int(mel_raw.shape[1]//3*2))
        # if random.random()>0.5:
        #     mel_refer = mel_raw[:,split:]
        # else:
        #     mel_refer = mel_raw[:,:split]
        # if mel_refer.shape[1]>200:
        #     mel_refer = mel_refer[:,:200]
        #text_token mel_codes 

        if mel_raw.shape[1]>400:
            mel_raw = mel_raw[:,:400]
            mel_codes = mel_codes[:100]
        if mel_codes.shape[-1]%2==1:
            mel_codes = mel_codes[:-1]
            mel_raw = mel_raw[:,:-4]
        return text_tokens, mel_codes, mel_raw, mel_refer

    def __len__(self):
        return len(self.audiopaths_and_text)


class DiffusionCollater():

    def __init__(self):
        pass
    def __call__(self, batch):
        batch = [x for x in batch if x is not None]
        if len(batch)==0:
            return None
        text_lens = [len(x[0]) for x in batch]
        max_text_len = max(text_lens)
        mel_code_lens = [len(x[1]) for x in batch]
        max_mel_code_len = max(mel_code_lens)
        mel_lens = [x[2].shape[1] for x in batch]
        max_mel_len = max(mel_lens)
        mel_refer_lens = [x[3].shape[1] for x in batch]
        max_mel_refer_len = max(mel_refer_lens)
        texts = []
        mel_codes = []
        mels = []
        mel_refers = []
        # This is the sequential "background" tokens that are used as padding for text tokens, as specified in the DALLE paper.
        for b in batch:
            text_token, mel_code, mel, mel_refer = b
            texts.append(F.pad(text_token,(0,max_text_len-len(text_token)), value=0))
            mel_codes.append(F.pad(mel_code,(0,max_mel_code_len-len(mel_code)), value=0))
            mels.append(F.pad(mel,(0, max_mel_len-mel.shape[1]), value=0))
            mel_refers.append(F.pad(mel_refer,(0, max_mel_refer_len-mel_refer.shape[1]), value=0))

        padded_text = torch.stack(texts)
        padded_mel_code = torch.stack(mel_codes)
        padded_mel = torch.stack(mels)
        padded_mel_refer = torch.stack(mel_refers)
        return {
            'padded_text': padded_text,
            'padded_mel_code': padded_mel_code,
            'padded_mel': padded_mel,
            'mel_lengths': LongTensor(mel_lens),
            'padded_mel_refer':padded_mel_refer,
            'mel_refer_lengths':LongTensor(mel_refer_lens)
        }


if __name__ == '__main__':
    params = {
        'mode': 'gpt_tts',
        'path': 'E:\\audio\\LJSpeech-1.1\\ljs_audio_text_train_filelist.txt',
        'phase': 'train',
        'n_workers': 0,
        'batch_size': 16,
        'mel_vocab_size': 512,
    }
    cfg = json.load(open('ttts/diffusion/config.json'))
    ds = DiffusionDataset(cfg)
    dl = torch.utils.data.DataLoader(ds, **cfg['dataloader'], collate_fn=DiffusionCollater())
    i = 0
    m = []
    max_text = 0
    max_mel = 0
    for b in tqdm(dl):
        break