File size: 1,821 Bytes
9a85c9a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
from transformers import AutoTokenizer, AutoModelForMaskedLM

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

tokenizer = AutoTokenizer.from_pretrained("./bert/chinese-roberta-wwm-ext-large")
model = AutoModelForMaskedLM.from_pretrained("./bert/chinese-roberta-wwm-ext-large").to(device)

def get_bert_feature(text, word2ph):
    with torch.no_grad():
        inputs = tokenizer(text, return_tensors='pt')
        for i in inputs:
            inputs[i] = inputs[i].to(device)
        res = model(**inputs, output_hidden_states=True)
        res = torch.cat(res['hidden_states'][-3:-2], -1)[0].cpu()

    assert len(word2ph) == len(text)+2
    word2phone = word2ph
    phone_level_feature = []
    for i in range(len(word2phone)):
        repeat_feature = res[i].repeat(word2phone[i], 1)
        phone_level_feature.append(repeat_feature)

    phone_level_feature = torch.cat(phone_level_feature, dim=0)


    return phone_level_feature.T

if __name__ == '__main__':
    # feature = get_bert_feature('你好,我是说的道理。')
    import torch

    word_level_feature = torch.rand(38, 1024)  # 12个词,每个词1024维特征
    word2phone = [1, 2, 1, 2, 2, 1, 2, 2, 1, 2, 2, 1, 2, 2, 2, 2, 2, 1, 1, 2, 2, 1, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 1]

    # 计算总帧数
    total_frames = sum(word2phone)
    print(word_level_feature.shape)
    print(word2phone)
    phone_level_feature = []
    for i in range(len(word2phone)):
        print(word_level_feature[i].shape)

        # 对每个词重复word2phone[i]次
        repeat_feature = word_level_feature[i].repeat(word2phone[i], 1)
        phone_level_feature.append(repeat_feature)

    phone_level_feature = torch.cat(phone_level_feature, dim=0)
    print(phone_level_feature.shape)  # torch.Size([36, 1024])