File size: 6,636 Bytes
f631117
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Copyright (c) 2024 Alibaba Inc (authors: Xiang Lyu)
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import torch

class CosyVoiceModel:

    def __init__(self,
                 llm: torch.nn.Module,
                 flow: torch.nn.Module,
                 hift: torch.nn.Module):
        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
        self.llm = llm
        self.flow = flow
        self.hift = hift

    def load(self, llm_model, flow_model, hift_model):
        self.llm.load_state_dict(torch.load(llm_model, map_location=self.device))
        self.llm.to(self.device).eval()
        self.flow.load_state_dict(torch.load(flow_model, map_location=self.device))
        self.flow.to(self.device).eval()
        self.hift.load_state_dict(torch.load(hift_model, map_location=self.device))
        self.hift.to(self.device).eval()

    def inference(self, text, text_len, flow_embedding, llm_embedding=torch.zeros(0, 192),
                  prompt_text=torch.zeros(1, 0, dtype=torch.int32), prompt_text_len=torch.zeros(1, dtype=torch.int32),
                  llm_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32), llm_prompt_speech_token_len=torch.zeros(1, dtype=torch.int32),
                  flow_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32), flow_prompt_speech_token_len=torch.zeros(1, dtype=torch.int32),
                  prompt_speech_feat=torch.zeros(1, 0, 80), prompt_speech_feat_len=torch.zeros(1, dtype=torch.int32)):
        tts_speech_token = self.llm.inference(text=text.to(self.device),
                                              text_len=text_len.to(self.device),
                                              prompt_text=prompt_text.to(self.device),
                                              prompt_text_len=prompt_text_len.to(self.device),
                                              prompt_speech_token=llm_prompt_speech_token.to(self.device),
                                              prompt_speech_token_len=llm_prompt_speech_token_len.to(self.device),
                                              embedding=llm_embedding.to(self.device),
                                              beam_size=1,
                                              sampling=25,
                                              max_token_text_ratio=30,
                                              min_token_text_ratio=3)
        tts_mel = self.flow.inference(token=tts_speech_token,
                                      token_len=torch.tensor([tts_speech_token.size(1)], dtype=torch.int32).to(self.device),
                                      prompt_token=flow_prompt_speech_token.to(self.device),
                                      prompt_token_len=flow_prompt_speech_token_len.to(self.device),
                                      prompt_feat=prompt_speech_feat.to(self.device),
                                      prompt_feat_len=prompt_speech_feat_len.to(self.device),
                                      embedding=flow_embedding.to(self.device))
        tts_speech = self.hift.inference(mel=tts_mel).cpu()
        torch.cuda.empty_cache()
        return {'tts_speech': tts_speech}

    def text_to_token(self, text, text_len, flow_embedding, llm_embedding=torch.zeros(0, 192),
                  prompt_text=torch.zeros(1, 0, dtype=torch.int32), prompt_text_len=torch.zeros(1, dtype=torch.int32),
                  llm_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32), llm_prompt_speech_token_len=torch.zeros(1, dtype=torch.int32),
                  flow_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32), flow_prompt_speech_token_len=torch.zeros(1, dtype=torch.int32),
                  prompt_speech_feat=torch.zeros(1, 0, 80), prompt_speech_feat_len=torch.zeros(1, dtype=torch.int32)):
        tts_speech_token = self.llm.inference(text=text.to(self.device),
                                              text_len=text_len.to(self.device),
                                              prompt_text=prompt_text.to(self.device),
                                              prompt_text_len=prompt_text_len.to(self.device),
                                              prompt_speech_token=llm_prompt_speech_token.to(self.device),
                                              prompt_speech_token_len=llm_prompt_speech_token_len.to(self.device),
                                              embedding=llm_embedding.to(self.device),
                                              beam_size=1,
                                              sampling=25,
                                              max_token_text_ratio=30,
                                              min_token_text_ratio=3)
        return tts_speech_token
        
    def token_to_speech(self, tts_speech_token, flow_embedding, llm_embedding=torch.zeros(0, 192),
                  prompt_text=torch.zeros(1, 0, dtype=torch.int32), prompt_text_len=torch.zeros(1, dtype=torch.int32),
                  llm_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32), llm_prompt_speech_token_len=torch.zeros(1, dtype=torch.int32),
                  flow_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32), flow_prompt_speech_token_len=torch.zeros(1, dtype=torch.int32),
                  prompt_speech_feat=torch.zeros(1, 0, 80), prompt_speech_feat_len=torch.zeros(1, dtype=torch.int32)):
        
        tts_mel = self.flow.inference(token=tts_speech_token,
                                      token_len=torch.tensor([tts_speech_token.size(1)], dtype=torch.int32).to(self.device),
                                      prompt_token=flow_prompt_speech_token.to(self.device),
                                      prompt_token_len=flow_prompt_speech_token_len.to(self.device),
                                      prompt_feat=prompt_speech_feat.to(self.device),
                                      prompt_feat_len=prompt_speech_feat_len.to(self.device),
                                      embedding=flow_embedding.to(self.device))
        tts_speech = self.hift.inference(mel=tts_mel).cpu()
        torch.cuda.empty_cache()
        return {'tts_speech': tts_speech}