cahya commited on
Commit
17f9ae1
1 Parent(s): 9fbe6a6

add api_token

Browse files
Files changed (1) hide show
  1. app.py +12 -11
app.py CHANGED
@@ -3,7 +3,7 @@ import torch
3
  from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
4
  import gradio as gr
5
  import sox
6
-
7
 
8
  def convert(inputfile, outfile):
9
  sox_tfm = sox.Transformer()
@@ -11,11 +11,12 @@ def convert(inputfile, outfile):
11
  file_type="wav", channels=1, encoding="signed-integer", rate=16000, bits=16
12
  )
13
  sox_tfm.build(inputfile, outfile)
14
-
15
 
 
 
16
  model_name = "indonesian-nlp/wav2vec2-indonesian-javanese-sundanese"
17
- processor = Wav2Vec2Processor.from_pretrained(model_name)
18
- model = Wav2Vec2ForCTC.from_pretrained(model_name)
19
 
20
  def parse_transcription(wav_file):
21
  filename = wav_file.name.split('.')[0]
@@ -30,16 +31,16 @@ def parse_transcription(wav_file):
30
 
31
 
32
  output = gr.outputs.Textbox(label="Indonesian, Javanese or Sundanese")
33
-
34
- input_ = gr.inputs.Audio(source="microphone", type="file")
35
- #gr.Interface(parse_transcription, inputs = input_, outputs="text",
36
- # analytics_enabled=False, show_tips=False, enable_queue=True).launch(inline=False);
37
-
38
- gr.Interface(parse_transcription, inputs = input_, outputs=[output],
39
  analytics_enabled=False,
40
  show_tips=False,
41
  theme='huggingface',
42
  layout='vertical',
43
  title="Multilingual Speech Recognition for Indonesian Languages",
44
- description="Speech Recognition Live Demo for Indonesian, Javanese and Sundanese",
 
 
45
  enable_queue=True).launch( inline=False)
 
3
  from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
4
  import gradio as gr
5
  import sox
6
+ import os
7
 
8
  def convert(inputfile, outfile):
9
  sox_tfm = sox.Transformer()
 
11
  file_type="wav", channels=1, encoding="signed-integer", rate=16000, bits=16
12
  )
13
  sox_tfm.build(inputfile, outfile)
 
14
 
15
+
16
+ api_token = os.getenv("API_TOKEN")
17
  model_name = "indonesian-nlp/wav2vec2-indonesian-javanese-sundanese"
18
+ processor = Wav2Vec2Processor.from_pretrained(model_name, use_auth_token=api_token)
19
+ model = Wav2Vec2ForCTC.from_pretrained(model_name, use_auth_token=api_token)
20
 
21
  def parse_transcription(wav_file):
22
  filename = wav_file.name.split('.')[0]
 
31
 
32
 
33
  output = gr.outputs.Textbox(label="Indonesian, Javanese or Sundanese")
34
+
35
+ input_ = gr.inputs.Audio(source="microphone", type="file")
36
+
37
+ gr.Interface(parse_transcription, inputs=input_, outputs=[output],
 
 
38
  analytics_enabled=False,
39
  show_tips=False,
40
  theme='huggingface',
41
  layout='vertical',
42
  title="Multilingual Speech Recognition for Indonesian Languages",
43
+ description="This is Speech Recognition live demo for Indonesian, Javanese and Sundanese Language. It uses"
44
+ "the Wav2Vec2 large model \"indonesian-nlp/wav2vec2-indonesian-javanese-sundanese\""
45
+ " fine-tuned on Common Voice and OpenSLR speech datasets.",
46
  enable_queue=True).launch( inline=False)