Spaces:
Running
on
Zero
Running
on
Zero
txya900619
commited on
Commit
•
f6c54a9
1
Parent(s):
a523a5e
fix: zero gpu bug
Browse files
app.py
CHANGED
@@ -46,7 +46,8 @@ OmegaConf.register_new_resolver("load_model", load_model)
|
|
46 |
models_config = OmegaConf.to_object(OmegaConf.load("configs/models.yaml"))
|
47 |
|
48 |
@spaces.GPU
|
49 |
-
def _do_tts(
|
|
|
50 |
if speaker_wav is not None:
|
51 |
return model.tts(
|
52 |
ipa,
|
@@ -70,7 +71,6 @@ def text_to_speech(
|
|
70 |
dialect,
|
71 |
text: str,
|
72 |
):
|
73 |
-
model = models_config[model_id]["model"]
|
74 |
if len(text) == 0:
|
75 |
raise gr.Error("請勿輸入空字串。")
|
76 |
words, ipa, pinyin, missing_words = get_ipa(text, dialect=dialect)
|
@@ -81,7 +81,7 @@ def text_to_speech(
|
|
81 |
parsed_ipa = parse_ipa(ipa)
|
82 |
if use_default_emb_or_custom == "default":
|
83 |
wav = _do_tts(
|
84 |
-
|
85 |
parsed_ipa,
|
86 |
speaker_name=speaker
|
87 |
if len(models_config[model_id]["speaker_mapping"]) > 1
|
@@ -91,7 +91,7 @@ def text_to_speech(
|
|
91 |
)
|
92 |
else:
|
93 |
wav = _do_tts(
|
94 |
-
|
95 |
parsed_ipa,
|
96 |
speaker_wav=speaker_wav,
|
97 |
language_name=dialect,
|
@@ -100,7 +100,7 @@ def text_to_speech(
|
|
100 |
return (
|
101 |
words,
|
102 |
pinyin,
|
103 |
-
(model.tts_model.config.audio.sample_rate, np.array(wav)),
|
104 |
)
|
105 |
|
106 |
|
|
|
46 |
models_config = OmegaConf.to_object(OmegaConf.load("configs/models.yaml"))
|
47 |
|
48 |
@spaces.GPU
|
49 |
+
def _do_tts(model_id, ipa, language_name,speaker_name=None, speaker_wav=None):
|
50 |
+
model = models_config[model_id]["model"]
|
51 |
if speaker_wav is not None:
|
52 |
return model.tts(
|
53 |
ipa,
|
|
|
71 |
dialect,
|
72 |
text: str,
|
73 |
):
|
|
|
74 |
if len(text) == 0:
|
75 |
raise gr.Error("請勿輸入空字串。")
|
76 |
words, ipa, pinyin, missing_words = get_ipa(text, dialect=dialect)
|
|
|
81 |
parsed_ipa = parse_ipa(ipa)
|
82 |
if use_default_emb_or_custom == "default":
|
83 |
wav = _do_tts(
|
84 |
+
model_id,
|
85 |
parsed_ipa,
|
86 |
speaker_name=speaker
|
87 |
if len(models_config[model_id]["speaker_mapping"]) > 1
|
|
|
91 |
)
|
92 |
else:
|
93 |
wav = _do_tts(
|
94 |
+
model_id,
|
95 |
parsed_ipa,
|
96 |
speaker_wav=speaker_wav,
|
97 |
language_name=dialect,
|
|
|
100 |
return (
|
101 |
words,
|
102 |
pinyin,
|
103 |
+
(models_config[model_id]["model"].tts_model.config.audio.sample_rate, np.array(wav)),
|
104 |
)
|
105 |
|
106 |
|