Pengwei Li commited on
Commit
8d99114
1 Parent(s): 3d50426

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +7 -57
app.py CHANGED
@@ -3,76 +3,26 @@ os.system("pip install gradio==2.8.0b2")
3
  import gradio as gr
4
  import numpy as np
5
 
6
- title = "fairseq S2S"
7
 
8
  description = "Gradio Demo for fairseq S2S: speech-to-speech translation models. To use it, simply add your audio, or click one of the examples to load them. Read more at the links below."
9
 
10
- article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2010.05171' target='_blank'>fairseq S2T: Fast Speech-to-Text Modeling with fairseq</a> | <a href='https://github.com/pytorch/fairseq/tree/main/examples/speech_to_text' target='_blank'>Github Repo</a></p>"
11
 
12
  examples = [
13
- ["common_voice_es_en.flac","xm_transformer_600m-es_en-multi_domain"],
14
- ["common_voice_ru_18945535.flac","xm_transformer_600m-ru_en-multi_domain"],
15
- ["common_voice_fr_19731305.mp3","xm_transformer_600m-fr_en-multi_domain"],
16
- ["common_voice_en_ru.mp3","xm_transformer_600m-en_ru-multi_domain"],
17
- ["common_voice_en_es.mp3","xm_transformer_600m-en_es-multi_domain"],
18
- ["common_voice_en_ar.mp3","xm_transformer_600m-en_ar-multi_domain"],
19
- ["common_voice_en_tr.mp3","xm_transformer_600m-en_tr-multi_domain"],
20
- ["common_voice_en_fr.mp3","xm_transformer_600m-en_fr-multi_domain"],
21
- ["common_voice_en_vi.mp3","xm_transformer_600m-en_vi-multi_domain"],
22
  ]
23
 
24
- io1 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-es_en-multi_domain")
25
-
26
- io2 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-ru_en-multi_domain")
27
-
28
- io3 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_ru-multi_domain")
29
-
30
- io4 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_es-multi_domain")
31
-
32
- io5 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_zh-multi_domain")
33
-
34
- io6 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-fr_en-multi_domain")
35
-
36
- io7 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_ar-multi_domain")
37
-
38
- io8 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_tr-multi_domain")
39
-
40
- io9 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_fr-multi_domain")
41
-
42
- io10 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_vi-multi_domain")
43
-
44
-
45
-
46
-
47
-
48
-
49
  def inference(text,model):
50
- if model == "xm_transformer_600m-es_en-multi_domain":
51
- outtext = io1(text)
52
- elif model == "xm_transformer_600m-ru_en-multi_domain":
53
- outtext = io2(text)
54
- elif model == "xm_transformer_600m-en_ru-multi_domain":
55
- outtext = io3(text)
56
- elif model == "xm_transformer_600m-en_es-multi_domain":
57
- outtext = io4(text)
58
- elif model == "xm_transformer_600m-en_zh-multi_domain":
59
- outtext = io5(text)
60
- elif model == "xm_transformer_600m-fr_en-multi_domain":
61
- outtext = io6(text)
62
- elif model == "xm_transformer_600m-en_ar-multi_domain":
63
- outtext = io7(text)
64
- elif model == "xm_transformer_600m-en_tr-multi_domain":
65
- outtext = io8(text)
66
- elif model == "xm_transformer_600m-en_fr-multi_domain":
67
- outtext = io9(text)
68
- else:
69
- outtext = io10(text)
70
  return outtext
71
 
72
 
73
  gr.Interface(
74
  inference,
75
- [gr.inputs.Audio(label="Input",type="filepath"),gr.inputs.Dropdown(choices=["xm_transformer_600m-es_en-multi_domain","xm_transformer_600m-ru_en-multi_domain","xm_transformer_600m-en_ru-multi_domain","xm_transformer_600m-en_es-multi_domain","xm_transformer_600m-fr_en-multi_domain","xm_transformer_600m-en_ar-multi_domain","xm_transformer_600m-en_tr-multi_domain","xm_transformer_600m-en_fr-multi_domain","xm_transformer_600m-en_vi-multi_domain"], type="value", default="xm_transformer_600m-es_en-multi_domain", label="model")
76
  ],
77
  gr.outputs.Audio(label="Output"),
78
  article=article,
 
3
  import gradio as gr
4
  import numpy as np
5
 
6
+ title = "Fairseq S2S"
7
 
8
  description = "Gradio Demo for fairseq S2S: speech-to-speech translation models. To use it, simply add your audio, or click one of the examples to load them. Read more at the links below."
9
 
10
+ article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2107.05604' target='_blank'>Direct speech-to-speech translation with discrete units</a> | <a href='https://github.com/facebookresearch/fairseq/tree/main/examples/speech_to_speech' target='_blank'>Github Repo</a></p>"
11
 
12
  examples = [
13
+ ["common_voice_es_en.flac","xm_transformer_s2ut_800m-es-en-st-asr-bt_h1_2022"],
 
 
 
 
 
 
 
 
14
  ]
15
 
16
+ io1 = gr.Interface.load("huggingface/facebook/xm_transformer_s2ut_800m-es-en-st-asr-bt_h1_2022")
17
+
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  def inference(text,model):
19
+ outtext = io1(text)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
20
  return outtext
21
 
22
 
23
  gr.Interface(
24
  inference,
25
+ [gr.inputs.Audio(label="Input",type="filepath"),gr.inputs.Dropdown(choices=["xm_transformer_s2ut_800m-es-en-st-asr-bt_h1_2022"], type="value", default="xm_transformer_s2ut_800m-es-en-st-asr-bt_h1_2022", label="model")
26
  ],
27
  gr.outputs.Audio(label="Output"),
28
  article=article,