lpw akhaliq HF staff commited on
Commit
4d64757
1 Parent(s): ac2b495

migrate to blocks from interface (#1)

Browse files

- migrate to blocks from interface (4c5e0a85807d2b0d947b016e572f97fb029da116)


Co-authored-by: Ahsen Khaliq <akhaliq@users.noreply.huggingface.co>

Files changed (1) hide show
  1. app.py +124 -23
app.py CHANGED
@@ -1,23 +1,12 @@
1
  import os
2
- os.system("pip install gradio==3.3")
3
  import gradio as gr
4
  import numpy as np
5
- import streamlit as st
6
 
7
- title = "Hokkien Translation"
8
 
9
- description = "Gradio Demo for fairseq speech-to-speech translation models. We have S2UT and UnitY models for bidirectional Hokkien and English translation. Please select the model and record your input to try it."
10
-
11
- article = "<p style='text-align: center'><a href='https://ai.facebook.com/blog/ai-translation-hokkien/' target='_blank'>Speech-to-speech translation for a real-world unwritten language</a> | <a href='https://github.com/facebookresearch/fairseq/tree/ust/examples/hokkien' target='_blank'>Github Repo</a></p>"
12
-
13
- examples = [
14
- ["enhanced_direct_s2st_units_audios_es-en_set2_source_12478_cv.flac","xm_transformer_s2ut_800m-es-en-st-asr-bt_h1_2022"],
15
- ]
16
-
17
- io1 = gr.Interface.load("huggingface/facebook/xm_transformer_s2ut_en-hk", api_key=st.secrets["api_key"])
18
- io2 = gr.Interface.load("huggingface/facebook/xm_transformer_s2ut_hk-en", api_key=st.secrets["api_key"])
19
- io3 = gr.Interface.load("huggingface/facebook/xm_transformer_unity_en-hk", api_key=st.secrets["api_key"])
20
- io4 = gr.Interface.load("huggingface/facebook/xm_transformer_unity_hk-en", api_key=st.secrets["api_key"])
21
 
22
  def inference(audio, model):
23
  if model == "xm_transformer_s2ut_en-hk":
@@ -30,12 +19,124 @@ def inference(audio, model):
30
  out_audio = io4(audio)
31
  return out_audio
32
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
 
34
- gr.Interface(
35
- inference,
36
- [gr.inputs.Audio(source="microphone", type="filepath", label="Input"),gr.inputs.Dropdown(choices=["xm_transformer_unity_en-hk", "xm_transformer_unity_hk-en", "xm_transformer_s2ut_en-hk", "xm_transformer_s2ut_hk-en"], default="xm_transformer_unity_en-hk",type="value", label="Model")
37
- ],
38
- gr.outputs.Audio(label="Output"),
39
- article=article,
40
- title=title,
41
- description=description).queue().launch()
 
1
  import os
 
2
  import gradio as gr
3
  import numpy as np
 
4
 
 
5
 
6
+ io1 = gr.Interface.load("huggingface/facebook/xm_transformer_s2ut_en-hk")
7
+ io2 = gr.Interface.load("huggingface/facebook/xm_transformer_s2ut_hk-en")
8
+ io3 = gr.Interface.load("huggingface/facebook/xm_transformer_unity_en-hk")
9
+ io4 = gr.Interface.load("huggingface/facebook/xm_transformer_unity_hk-en")
 
 
 
 
 
 
 
 
10
 
11
  def inference(audio, model):
12
  if model == "xm_transformer_s2ut_en-hk":
 
19
  out_audio = io4(audio)
20
  return out_audio
21
 
22
+
23
+ css = """
24
+ .gradio-container {
25
+ font-family: 'IBM Plex Sans', sans-serif;
26
+ }
27
+ .gr-button {
28
+ color: white;
29
+ border-color: black;
30
+ background: black;
31
+ }
32
+ input[type='range'] {
33
+ accent-color: black;
34
+ }
35
+ .dark input[type='range'] {
36
+ accent-color: #dfdfdf;
37
+ }
38
+ .container {
39
+ max-width: 730px;
40
+ margin: auto;
41
+ padding-top: 1.5rem;
42
+ }
43
+
44
+ .details:hover {
45
+ text-decoration: underline;
46
+ }
47
+ .gr-button {
48
+ white-space: nowrap;
49
+ }
50
+ .gr-button:focus {
51
+ border-color: rgb(147 197 253 / var(--tw-border-opacity));
52
+ outline: none;
53
+ box-shadow: var(--tw-ring-offset-shadow), var(--tw-ring-shadow), var(--tw-shadow, 0 0 #0000);
54
+ --tw-border-opacity: 1;
55
+ --tw-ring-offset-shadow: var(--tw-ring-inset) 0 0 0 var(--tw-ring-offset-width) var(--tw-ring-offset-color);
56
+ --tw-ring-shadow: var(--tw-ring-inset) 0 0 0 calc(3px var(--tw-ring-offset-width)) var(--tw-ring-color);
57
+ --tw-ring-color: rgb(191 219 254 / var(--tw-ring-opacity));
58
+ --tw-ring-opacity: .5;
59
+ }
60
+ .footer {
61
+ margin-bottom: 45px;
62
+ margin-top: 35px;
63
+ text-align: center;
64
+ border-bottom: 1px solid #e5e5e5;
65
+ }
66
+ .footer>p {
67
+ font-size: .8rem;
68
+ display: inline-block;
69
+ padding: 0 10px;
70
+ transform: translateY(10px);
71
+ background: white;
72
+ }
73
+ .dark .footer {
74
+ border-color: #303030;
75
+ }
76
+ .dark .footer>p {
77
+ background: #0b0f19;
78
+ }
79
+ .prompt h4{
80
+ margin: 1.25em 0 .25em 0;
81
+ font-weight: bold;
82
+ font-size: 115%;
83
+ }
84
+ .animate-spin {
85
+ animation: spin 1s linear infinite;
86
+ }
87
+ @keyframes spin {
88
+ from {
89
+ transform: rotate(0deg);
90
+ }
91
+ to {
92
+ transform: rotate(360deg);
93
+ }
94
+ }
95
+ """
96
+
97
+ block = gr.Blocks(css=css)
98
+
99
+
100
+
101
+ with block:
102
+ gr.HTML(
103
+ """
104
+ <div style="text-align: center; max-width: 650px; margin: 0 auto;">
105
+ <div
106
+ style="
107
+ display: inline-flex;
108
+ align-items: center;
109
+ gap: 0.8rem;
110
+ font-size: 1.75rem;
111
+ "
112
+ >
113
+ <h1 style="font-weight: 900; margin-bottom: 7px;">
114
+ Hokkien Translation
115
+ </h1>
116
+ </div>
117
+ <p style="margin-bottom: 10px; font-size: 94%">
118
+ Gradio Demo for fairseq speech-to-speech translation models. We have S2UT and UnitY models for bidirectional Hokkien and English translation. Please select the model and record your input to try it.
119
+ </p>
120
+ </div>
121
+ """
122
+ )
123
+ with gr.Group():
124
+ with gr.Box():
125
+ with gr.Row().style(mobile_collapse=False, equal_height=True):
126
+ audio = gr.Audio(
127
+ source="microphone", type="filepath", label="Input"
128
+ )
129
+
130
+ btn = gr.Button("Submit")
131
+ model = gr.Dropdown(choices=["xm_transformer_unity_en-hk", "xm_transformer_unity_hk-en", "xm_transformer_s2ut_en-hk", "xm_transformer_s2ut_hk-en"], value="xm_transformer_unity_en-hk",type="value", label="Model")
132
+ out = gr.Audio(label="Output")
133
+
134
+ btn.click(inference, inputs=[audio, model], outputs=out)
135
+ gr.HTML('''
136
+ <div class="footer">
137
+ <p>Model by <a href="https://ai.facebook.com/" style="text-decoration: underline;" target="_blank">Meta AI</a>
138
+ </p>
139
+ </div>
140
+ ''')
141
 
142
+ block.launch()