Spaces:
Running
on
Zero
Running
on
Zero
set onnx to false as last chunk rtf unstable
Browse files- .github/workflows/lint.yml +1 -0
- cosyvoice/cli/cosyvoice.py +1 -1
- cosyvoice/cli/model.py +43 -50
.github/workflows/lint.yml
CHANGED
@@ -2,6 +2,7 @@ name: Lint
|
|
2 |
|
3 |
on:
|
4 |
pull_request:
|
|
|
5 |
|
6 |
jobs:
|
7 |
quick-checks:
|
|
|
2 |
|
3 |
on:
|
4 |
pull_request:
|
5 |
+
push:
|
6 |
|
7 |
jobs:
|
8 |
quick-checks:
|
cosyvoice/cli/cosyvoice.py
CHANGED
@@ -23,7 +23,7 @@ from cosyvoice.utils.file_utils import logging
|
|
23 |
|
24 |
class CosyVoice:
|
25 |
|
26 |
-
def __init__(self, model_dir, load_jit=True, load_onnx=
|
27 |
instruct = True if '-Instruct' in model_dir else False
|
28 |
self.model_dir = model_dir
|
29 |
if not os.path.exists(model_dir):
|
|
|
23 |
|
24 |
class CosyVoice:
|
25 |
|
26 |
+
def __init__(self, model_dir, load_jit=True, load_onnx=False):
|
27 |
instruct = True if '-Instruct' in model_dir else False
|
28 |
self.model_dir = model_dir
|
29 |
if not os.path.exists(model_dir):
|
cosyvoice/cli/model.py
CHANGED
@@ -43,7 +43,6 @@ class CosyVoiceModel:
|
|
43 |
self.stream_scale_factor = 1
|
44 |
assert self.stream_scale_factor >= 1, 'stream_scale_factor should be greater than 1, change it according to your actual rtf'
|
45 |
self.llm_context = torch.cuda.stream(torch.cuda.Stream(self.device)) if torch.cuda.is_available() else nullcontext()
|
46 |
-
self.flow_hift_context = torch.cuda.stream(torch.cuda.Stream(self.device)) if torch.cuda.is_available() else nullcontext()
|
47 |
self.lock = threading.Lock()
|
48 |
# dict used to store session related variable
|
49 |
self.tts_speech_token_dict = {}
|
@@ -93,32 +92,31 @@ class CosyVoiceModel:
|
|
93 |
self.llm_end_dict[uuid] = True
|
94 |
|
95 |
def token2wav(self, token, prompt_token, prompt_feat, embedding, uuid, finalize=False):
|
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 |
-
tts_speech, tts_source = self.hift.inference(mel=tts_mel, cache_source=hift_cache_source)
|
122 |
return tts_speech
|
123 |
|
124 |
def inference(self, text, flow_embedding, llm_embedding=torch.zeros(0, 192),
|
@@ -139,13 +137,12 @@ class CosyVoiceModel:
|
|
139 |
time.sleep(0.1)
|
140 |
if len(self.tts_speech_token_dict[this_uuid]) >= token_hop_len + self.token_overlap_len:
|
141 |
this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid][:token_hop_len + self.token_overlap_len], dim=1)
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
finalize=False)
|
149 |
yield {'tts_speech': this_tts_speech.cpu()}
|
150 |
with self.lock:
|
151 |
self.tts_speech_token_dict[this_uuid] = self.tts_speech_token_dict[this_uuid][token_hop_len:]
|
@@ -156,30 +153,26 @@ class CosyVoiceModel:
|
|
156 |
p.join()
|
157 |
# deal with remain tokens, make sure inference remain token len equals token_hop_len when cache_speech is not None
|
158 |
this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid], dim=1)
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
finalize=True)
|
166 |
yield {'tts_speech': this_tts_speech.cpu()}
|
167 |
else:
|
168 |
# deal with all tokens
|
169 |
p.join()
|
170 |
this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid], dim=1)
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
finalize=True)
|
178 |
yield {'tts_speech': this_tts_speech.cpu()}
|
179 |
with self.lock:
|
180 |
self.tts_speech_token_dict.pop(this_uuid)
|
181 |
self.llm_end_dict.pop(this_uuid)
|
182 |
self.mel_overlap_dict.pop(this_uuid)
|
183 |
self.hift_cache_dict.pop(this_uuid)
|
184 |
-
if torch.cuda.is_available():
|
185 |
-
torch.cuda.synchronize()
|
|
|
43 |
self.stream_scale_factor = 1
|
44 |
assert self.stream_scale_factor >= 1, 'stream_scale_factor should be greater than 1, change it according to your actual rtf'
|
45 |
self.llm_context = torch.cuda.stream(torch.cuda.Stream(self.device)) if torch.cuda.is_available() else nullcontext()
|
|
|
46 |
self.lock = threading.Lock()
|
47 |
# dict used to store session related variable
|
48 |
self.tts_speech_token_dict = {}
|
|
|
92 |
self.llm_end_dict[uuid] = True
|
93 |
|
94 |
def token2wav(self, token, prompt_token, prompt_feat, embedding, uuid, finalize=False):
|
95 |
+
tts_mel = self.flow.inference(token=token.to(self.device),
|
96 |
+
token_len=torch.tensor([token.shape[1]], dtype=torch.int32).to(self.device),
|
97 |
+
prompt_token=prompt_token.to(self.device),
|
98 |
+
prompt_token_len=torch.tensor([prompt_token.shape[1]], dtype=torch.int32).to(self.device),
|
99 |
+
prompt_feat=prompt_feat.to(self.device),
|
100 |
+
prompt_feat_len=torch.tensor([prompt_feat.shape[1]], dtype=torch.int32).to(self.device),
|
101 |
+
embedding=embedding.to(self.device))
|
102 |
+
# mel overlap fade in out
|
103 |
+
if self.mel_overlap_dict[uuid] is not None:
|
104 |
+
tts_mel = fade_in_out(tts_mel, self.mel_overlap_dict[uuid], self.mel_window)
|
105 |
+
# append hift cache
|
106 |
+
if self.hift_cache_dict[uuid] is not None:
|
107 |
+
hift_cache_mel, hift_cache_source = self.hift_cache_dict[uuid]['mel'], self.hift_cache_dict[uuid]['source']
|
108 |
+
tts_mel = torch.concat([hift_cache_mel, tts_mel], dim=2)
|
109 |
+
else:
|
110 |
+
hift_cache_source = torch.zeros(1, 1, 0)
|
111 |
+
# keep overlap mel and hift cache
|
112 |
+
if finalize is False:
|
113 |
+
self.mel_overlap_dict[uuid] = tts_mel[:, :, -self.mel_overlap_len:]
|
114 |
+
tts_mel = tts_mel[:, :, :-self.mel_overlap_len]
|
115 |
+
tts_speech, tts_source = self.hift.inference(mel=tts_mel, cache_source=hift_cache_source)
|
116 |
+
self.hift_cache_dict[uuid] = {'source': tts_source[:, :, -self.source_cache_len:], 'mel': tts_mel[:, :, -self.mel_cache_len:]}
|
117 |
+
tts_speech = tts_speech[:, :-self.source_cache_len]
|
118 |
+
else:
|
119 |
+
tts_speech, tts_source = self.hift.inference(mel=tts_mel, cache_source=hift_cache_source)
|
|
|
120 |
return tts_speech
|
121 |
|
122 |
def inference(self, text, flow_embedding, llm_embedding=torch.zeros(0, 192),
|
|
|
137 |
time.sleep(0.1)
|
138 |
if len(self.tts_speech_token_dict[this_uuid]) >= token_hop_len + self.token_overlap_len:
|
139 |
this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid][:token_hop_len + self.token_overlap_len], dim=1)
|
140 |
+
this_tts_speech = self.token2wav(token=this_tts_speech_token,
|
141 |
+
prompt_token=flow_prompt_speech_token,
|
142 |
+
prompt_feat=prompt_speech_feat,
|
143 |
+
embedding=flow_embedding,
|
144 |
+
uuid=this_uuid,
|
145 |
+
finalize=False)
|
|
|
146 |
yield {'tts_speech': this_tts_speech.cpu()}
|
147 |
with self.lock:
|
148 |
self.tts_speech_token_dict[this_uuid] = self.tts_speech_token_dict[this_uuid][token_hop_len:]
|
|
|
153 |
p.join()
|
154 |
# deal with remain tokens, make sure inference remain token len equals token_hop_len when cache_speech is not None
|
155 |
this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid], dim=1)
|
156 |
+
this_tts_speech = self.token2wav(token=this_tts_speech_token,
|
157 |
+
prompt_token=flow_prompt_speech_token,
|
158 |
+
prompt_feat=prompt_speech_feat,
|
159 |
+
embedding=flow_embedding,
|
160 |
+
uuid=this_uuid,
|
161 |
+
finalize=True)
|
|
|
162 |
yield {'tts_speech': this_tts_speech.cpu()}
|
163 |
else:
|
164 |
# deal with all tokens
|
165 |
p.join()
|
166 |
this_tts_speech_token = torch.concat(self.tts_speech_token_dict[this_uuid], dim=1)
|
167 |
+
this_tts_speech = self.token2wav(token=this_tts_speech_token,
|
168 |
+
prompt_token=flow_prompt_speech_token,
|
169 |
+
prompt_feat=prompt_speech_feat,
|
170 |
+
embedding=flow_embedding,
|
171 |
+
uuid=this_uuid,
|
172 |
+
finalize=True)
|
|
|
173 |
yield {'tts_speech': this_tts_speech.cpu()}
|
174 |
with self.lock:
|
175 |
self.tts_speech_token_dict.pop(this_uuid)
|
176 |
self.llm_end_dict.pop(this_uuid)
|
177 |
self.mel_overlap_dict.pop(this_uuid)
|
178 |
self.hift_cache_dict.pop(this_uuid)
|
|
|
|