lengyue233 commited on
Commit
45eadd5
1 Parent(s): 9bd8a0b

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +27 -33
app.py CHANGED
@@ -119,41 +119,19 @@ def build_html_error_message(error):
119
  @GPU_DECORATOR
120
  @torch.inference_mode()
121
  def inference(req: ServeTTSRequest):
122
-
123
- idstr: str | None = req.reference_id
124
- if idstr is not None:
125
- ref_folder = Path("references") / idstr
126
- ref_folder.mkdir(parents=True, exist_ok=True)
127
- ref_audios = list_files(
128
- ref_folder, AUDIO_EXTENSIONS, recursive=True, sort=False
 
129
  )
 
 
 
130
 
131
- prompt_tokens = [
132
- encode_reference(
133
- decoder_model=decoder_model,
134
- reference_audio=audio_to_bytes(str(ref_audio)),
135
- enable_reference_audio=True,
136
- )
137
- for ref_audio in ref_audios
138
- ]
139
- prompt_texts = [
140
- read_ref_text(str(ref_audio.with_suffix(".lab")))
141
- for ref_audio in ref_audios
142
- ]
143
-
144
- else:
145
- # Parse reference audio aka prompt
146
- refs = req.references
147
-
148
- prompt_tokens = [
149
- encode_reference(
150
- decoder_model=decoder_model,
151
- reference_audio=ref.audio,
152
- enable_reference_audio=True,
153
- )
154
- for ref in refs
155
- ]
156
- prompt_texts = [ref.text for ref in refs]
157
  if req.seed is not None:
158
  set_seed(req.seed)
159
  logger.warning(f"set seed: {req.seed}")
@@ -411,11 +389,27 @@ def build_app():
411
  seed,
412
  use_memory_cache,
413
  ):
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
414
  references = []
415
  if reference_audio:
416
  # 将文件路径转换为字节
417
  with open(reference_audio, 'rb') as audio_file:
418
  audio_bytes = audio_file.read()
 
419
  references = [
420
  ServeReferenceAudio(audio=audio_bytes, text=reference_text)
421
  ]
 
119
  @GPU_DECORATOR
120
  @torch.inference_mode()
121
  def inference(req: ServeTTSRequest):
122
+ # Parse reference audio aka prompt
123
+ refs = req.references
124
+
125
+ prompt_tokens = [
126
+ encode_reference(
127
+ decoder_model=decoder_model,
128
+ reference_audio=ref.audio,
129
+ enable_reference_audio=True,
130
  )
131
+ for ref in refs
132
+ ]
133
+ prompt_texts = [ref.text for ref in refs]
134
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
135
  if req.seed is not None:
136
  set_seed(req.seed)
137
  logger.warning(f"set seed: {req.seed}")
 
389
  seed,
390
  use_memory_cache,
391
  ):
392
+ print(
393
+ "call inference wrapper",
394
+ text,
395
+ normalize,
396
+ reference_audio,
397
+ reference_text,
398
+ max_new_tokens,
399
+ chunk_length,
400
+ top_p,
401
+ repetition_penalty,
402
+ temperature,
403
+ seed,
404
+ use_memory_cache
405
+ )
406
+
407
  references = []
408
  if reference_audio:
409
  # 将文件路径转换为字节
410
  with open(reference_audio, 'rb') as audio_file:
411
  audio_bytes = audio_file.read()
412
+
413
  references = [
414
  ServeReferenceAudio(audio=audio_bytes, text=reference_text)
415
  ]