File size: 5,930 Bytes
c5ed230
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7a58c1e
c5ed230
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5854014
c5ed230
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5854014
c5ed230
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5854014
c5ed230
 
 
 
 
 
 
 
 
 
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
import os
import json
import logging
import config
import numpy as np
from utils.utils import check_is_none
from voice import vits, TTS

lang_dict = {
    "english_cleaners": ["en"],
    "english_cleaners2": ["en"],
    "japanese_cleaners": ["ja"],
    "japanese_cleaners2": ["ja"],
    "korean_cleaners": ["ko"],
    "chinese_cleaners": ["zh"],
    "zh_ja_mixture_cleaners": ["zh", "ja"],
    "sanskrit_cleaners": ["sa"],
    "cjks_cleaners": ["zh", "ja", "ko", "sa"],
    "cjke_cleaners": ["zh", "ja", "ko", "en"],
    "cjke_cleaners2": ["zh", "ja", "ko", "en"],
    "cje_cleaners": ["zh", "ja", "en"],
    "thai_cleaners": ["th"],
    "shanghainese_cleaners": ["sh"],
    "chinese_dialect_cleaners": ["zh", "ja", "sh", "gd", "en", "SZ", "WX", "CZ", "HZ", "SX", "NB", "JJ", "YX", "JD",
                                 "ZR", "PH", "TX", "JS", "HN", "LP", "XS", "FY", "RA", "CX", "SM", "TT", "WZ", "SC",
                                 "YB"],
}


def analysis(model_config_json):
    model_config = json.load(model_config_json)
    symbols = model_config.get("symbols", None)
    emotion_embedding = model_config.get("data").get("emotion_embedding", False)
    if symbols != None:
        if not emotion_embedding:
            mode_type = "vits"
        else:
            mode_type = "w2v2"
    else:
        mode_type = "hubert"
    return mode_type


def load_npy(model_):
    if isinstance(model_, list):
        # check if is .npy
        for i in model_:
            _model_extention = os.path.splitext(i)[1]
            if _model_extention != ".npy":
                raise ValueError(f"Unsupported model type: {_model_extention}")

        # merge npy files
        emotion_reference = np.empty((0, 1024))
        for i in model_:
            tmp = np.load(i).reshape(-1, 1024)
            emotion_reference = np.append(emotion_reference, tmp, axis=0)

    elif os.path.isdir(model_):
        emotion_reference = np.empty((0, 1024))
        for root, dirs, files in os.walk(model_):
            for file_name in files:
                # check if is .npy
                _model_extention = os.path.splitext(file_name)[1]
                if _model_extention != ".npy":
                    continue
                file_path = os.path.join(root, file_name)

                # merge npy files
                tmp = np.load(file_path).reshape(-1, 1024)
                emotion_reference = np.append(emotion_reference, tmp, axis=0)

    elif os.path.isfile(model_):
        # check if is .npy
        _model_extention = os.path.splitext(model_)[1]
        if _model_extention != ".npy":
            raise ValueError(f"Unsupported model type: {_model_extention}")

        emotion_reference = np.load(model_)
    logging.info(f"Loaded emotional dimention npy range:{len(emotion_reference)}")
    return emotion_reference


def merge_model(merging_model):
    vits_obj = []
    vits_speakers = []
    hubert_vits_obj = []
    hubert_vits_speakers = []
    w2v2_vits_obj = []
    w2v2_vits_speakers = []

    # model list
    vits_list = []
    hubert_vits_list = []
    w2v2_vits_list = []

    for l in merging_model:
        with open(l[1], 'r', encoding='utf-8') as model_config:
            model_type = analysis(model_config)
        if model_type == "vits":
            vits_list.append(l)
        elif model_type == "hubert":
            hubert_vits_list.append(l)
        elif model_type == "w2v2":
            w2v2_vits_list.append(l)

    # merge vits
    new_id = 0
    for obj_id, i in enumerate(vits_list):
        obj = vits(model=i[0], config=i[1], model_type="vits")
        lang = lang_dict.get(obj.get_cleaner(), obj.get_cleaner())

        for id, name in enumerate(obj.get_speakers()):
            vits_obj.append([int(id), obj, obj_id])
            vits_speakers.append({"id": new_id, "name": name, "lang": lang})
            new_id += 1

    # merge hubert-vits
    if len(hubert_vits_list) != 0:
        if getattr(config, "HUBERT_SOFT_MODEL", None) == None or check_is_none(config.HUBERT_SOFT_MODEL):
            raise ValueError(f"Please configure HUBERT_SOFT_MODEL path in config.py")
        try:
            from hubert_model import hubert_soft
            hubert = hubert_soft(config.HUBERT_SOFT_MODEL)
        except Exception as e:
            raise ValueError(f"Load HUBERT_SOFT_MODEL failed {e}")

    new_id = 0
    for obj_id, i in enumerate(hubert_vits_list):
        obj = vits(model=i[0], config=i[1], model_=hubert, model_type="hubert")
        lang = lang_dict.get(obj.get_cleaner(), obj.get_cleaner())

        for id, name in enumerate(obj.get_speakers()):
            hubert_vits_obj.append([int(id), obj, obj_id])
            hubert_vits_speakers.append({"id": new_id, "name": name, "lang": lang})
            new_id += 1

    # merge w2v2-vits
    if len(w2v2_vits_list) != 0:
        if getattr(config, "DIMENSIONAL_EMOTION_NPY", None) == None or check_is_none(config.DIMENSIONAL_EMOTION_NPY):
            raise ValueError(f"Please configure DIMENSIONAL_EMOTION_NPY path in config.py")
        try:
            emotion_reference = load_npy(config.DIMENSIONAL_EMOTION_NPY)
        except Exception as e:
            raise ValueError(f"Load DIMENSIONAL_EMOTION_NPY failed {e}")

    new_id = 0
    for obj_id, i in enumerate(w2v2_vits_list):
        obj = vits(model=i[0], config=i[1], model_=emotion_reference, model_type="w2v2")
        lang = lang_dict.get(obj.get_cleaner(), obj.get_cleaner())

        for id, name in enumerate(obj.get_speakers()):
            w2v2_vits_obj.append([int(id), obj, obj_id])
            w2v2_vits_speakers.append({"id": new_id, "name": name, "lang": lang})
            new_id += 1

    voice_obj = {"VITS": vits_obj, "HUBERT-VITS": hubert_vits_obj, "W2V2-VITS": w2v2_vits_obj}
    voice_speakers = {"VITS": vits_speakers, "HUBERT-VITS": hubert_vits_speakers, "W2V2-VITS": w2v2_vits_speakers}

    tts = TTS(voice_obj, voice_speakers)

    return tts