bomolopuu commited on
Commit
d3bad75
·
1 Parent(s): ad2000a

change .name to obj as is

Browse files
Files changed (1) hide show
  1. asr.py +5 -2
asr.py CHANGED
@@ -6,6 +6,7 @@ from pathlib import Path
6
  import torch
7
  from transformers import Wav2Vec2ForCTC, AutoProcessor
8
  import numpy as np
 
9
 
10
  # Настройка логирования
11
  logging.basicConfig(level=logging.DEBUG)
@@ -26,10 +27,10 @@ model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
26
 
27
  def safe_process_file(file_obj):
28
  try:
29
- logger.debug(f"Processing file: {file_obj.name}")
30
 
31
  # Используем Path для безопасной обработки путей
32
- file_path = Path(file_obj.name)
33
 
34
  logger.debug(f"Loading audio from file path: {file_path}")
35
 
@@ -61,6 +62,7 @@ def transcribe_multiple_files(audio_files, lang, transcription):
61
  transcriptions.append(f"Error processing file: {str(e)}\n")
62
  return "\n".join(transcriptions)
63
 
 
64
  def transcribe_file(model, audio_samples, lang, user_transcription):
65
  if not audio_samples:
66
  return "<<ERROR: Empty Audio Input>>"
@@ -92,6 +94,7 @@ def transcribe_file(model, audio_samples, lang, user_transcription):
92
 
93
  #return transcription
94
 
 
95
  def fine_tune_model(model, processor, user_transcription, audio_samples, lang_code):
96
  # Convert the user-provided transcription to a tensor
97
  transcription_tensor = processor.tokenize(user_transcription, return_tensors="pt")
 
6
  import torch
7
  from transformers import Wav2Vec2ForCTC, AutoProcessor
8
  import numpy as np
9
+ import spaces
10
 
11
  # Настройка логирования
12
  logging.basicConfig(level=logging.DEBUG)
 
27
 
28
  def safe_process_file(file_obj):
29
  try:
30
+ logger.debug(f"Processing file: {file_obj}")
31
 
32
  # Используем Path для безопасной обработки путей
33
+ file_path = Path(file_obj)
34
 
35
  logger.debug(f"Loading audio from file path: {file_path}")
36
 
 
62
  transcriptions.append(f"Error processing file: {str(e)}\n")
63
  return "\n".join(transcriptions)
64
 
65
+ @spaces.GPU
66
  def transcribe_file(model, audio_samples, lang, user_transcription):
67
  if not audio_samples:
68
  return "<<ERROR: Empty Audio Input>>"
 
94
 
95
  #return transcription
96
 
97
+ @spaces.GPU
98
  def fine_tune_model(model, processor, user_transcription, audio_samples, lang_code):
99
  # Convert the user-provided transcription to a tensor
100
  transcription_tensor = processor.tokenize(user_transcription, return_tensors="pt")