csukuangfj
commited on
Commit
•
3a556c6
1
Parent(s):
fb72693
html for kws
Browse files- generate-asr-2pass.py +34 -2
- generate-asr.py +7 -13
- generate-kws.py +158 -0
- generate-vad-asr.py +169 -0
- generate-vad.py +149 -0
- kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-arm64-v8a.apk +0 -3
- kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-armeabi-v7a.apk +0 -3
- kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-x86.apk +0 -3
- kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-x86_64.apk +0 -3
- kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-arm64-v8a.apk +0 -3
- kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-armeabi-v7a.apk +0 -3
- kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-x86.apk +0 -3
- kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-x86_64.apk +0 -3
generate-asr-2pass.py
CHANGED
@@ -61,8 +61,8 @@ def get_all_files(d: str, suffix: str) -> List[str]:
|
|
61 |
|
62 |
def to_file(filename: str, files: List[str]):
|
63 |
content = r"""
|
64 |
-
<h1> APKs for
|
65 |
-
This page lists the <strong>
|
66 |
one of the deployment frameworks of <a href="https://github.com/k2-fsa">the Next-gen Kaldi project</a>.
|
67 |
<br/>
|
68 |
The name of an APK has the following rule:
|
@@ -79,6 +79,38 @@ where
|
|
79 |
|
80 |
<br/>
|
81 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
82 |
You can download all supported models from
|
83 |
<a href="https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models">https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models</a>
|
84 |
|
|
|
61 |
|
62 |
def to_file(filename: str, files: List[str]):
|
63 |
content = r"""
|
64 |
+
<h1> APKs for two-pass speech recognition </h1>
|
65 |
+
This page lists the <strong>two-pass speech recognition</strong> APKs for <a href="http://github.com/k2-fsa/sherpa-onnx">sherpa-onnx</a>,
|
66 |
one of the deployment frameworks of <a href="https://github.com/k2-fsa">the Next-gen Kaldi project</a>.
|
67 |
<br/>
|
68 |
The name of an APK has the following rule:
|
|
|
79 |
|
80 |
<br/>
|
81 |
|
82 |
+
<span style="color:red;">Two-pass:</span> Two models are used during the recognition. A streaming model
|
83 |
+
is used in the first pass, while a non-streaming model is used in the second pass. The purpose
|
84 |
+
of the first pass is to give feedback to users that the system is working and it displays results while
|
85 |
+
users are speaking. When an endpoint is detected, the samples between two endpoints are sent to the second
|
86 |
+
pass model for recognition. The output of the second pass model is the final recognition result.
|
87 |
+
|
88 |
+
<br/>
|
89 |
+
<br/>
|
90 |
+
|
91 |
+
<span style="color:red;">Features of the first-pass model:</span>
|
92 |
+
<ul>
|
93 |
+
<li> A streaming model
|
94 |
+
<li> Low latency
|
95 |
+
<li> Small in model size
|
96 |
+
<li> Runs very fast
|
97 |
+
<li> Low accuracy
|
98 |
+
</ul>
|
99 |
+
|
100 |
+
<br/>
|
101 |
+
|
102 |
+
<span style="color:red;">Features of the second-pass model:</span>
|
103 |
+
<ul>
|
104 |
+
<li> A non-streaming model
|
105 |
+
<li> High latency
|
106 |
+
<li> Large in model size
|
107 |
+
<li> Runs a bit slow
|
108 |
+
<li> High accuracy
|
109 |
+
</ul>
|
110 |
+
|
111 |
+
<br/>
|
112 |
+
<br/>
|
113 |
+
|
114 |
You can download all supported models from
|
115 |
<a href="https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models">https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models</a>
|
116 |
|
generate-asr.py
CHANGED
@@ -103,40 +103,34 @@ see https://www.tablesgenerator.com/html_tables#
|
|
103 |
<tr>
|
104 |
<th class="tg-0pky">APK</th>
|
105 |
<th class="tg-0lax">Comment</th>
|
106 |
-
<th class="tg-0pky">Model
|
107 |
-
<th class="tg-0pky">Model Link</th>
|
108 |
</tr>
|
109 |
</thead>
|
110 |
<tbody>
|
111 |
<tr>
|
112 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-bilingual_zh_en-zipformer.apk</td>
|
113 |
<td class="tg-0lax">It supports both English and Chinese.</td>
|
114 |
-
<td class="tg-0pky">sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20</td>
|
115 |
-
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2">https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2</a></td>
|
116 |
</tr>
|
117 |
<tr>
|
118 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-fr-zipformer.apk</td>
|
119 |
<td class="tg-0lax"><span style="font-weight:400;font-style:normal">It supports only French.</span></td>
|
120 |
-
<td class="tg-0pky">sherpa-onnx-streaming-zipformer-fr-2023-04-14</td>
|
121 |
-
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-fr-2023-04-14.tar.bz2">https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-fr-2023-04-14.tar.bz2</a></td>
|
122 |
</tr>
|
123 |
<tr>
|
124 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-zh-zipformer2.apk</td>
|
125 |
<td class="tg-0lax">It supports only Chinese.</td>
|
126 |
-
<td class="tg-0pky">icefall-asr-zipformer-streaming-wenetspeech-20230615</td>
|
127 |
-
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/icefall-asr-zipformer-streaming-wenetspeech-20230615.tar.bz2">https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/icefall-asr-zipformer-streaming-wenetspeech-20230615.tar.bz2</a></td>
|
128 |
</tr>
|
129 |
<tr>
|
130 |
-
<td class="tg-0pky">sherpa-onnx-x.y.
|
131 |
<td class="tg-0lax">It supports only Chinese.<br>It uses the smallest zipformer and runs<br>super fast, though its accuracy is not that good.</td>
|
132 |
-
<td class="tg-0pky">sherpa-onnx-streaming-zipformer-zh-14M-2023-02-23</td>
|
133 |
-
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-zh-14M-2023-02-23.tar.bz2">https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-zh-14M-2023-02-23.tar.bz2</a></td>
|
134 |
</tr>
|
135 |
<tr>
|
136 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-en-small_zipformer.apk</td>
|
137 |
<td class="tg-0lax">It suppors only English.<br>It uses a very small zipformer and runs<br>super fast, though its accuracy is not that good.</td>
|
138 |
-
<td class="tg-0pky">sherpa-onnx-streaming-zipformer-en-20M-2023-02-17</td>
|
139 |
-
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-en-20M-2023-02-17.tar.bz2">https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-en-20M-2023-02-17.tar.bz2</a></td>
|
140 |
</tr>
|
141 |
</tbody>
|
142 |
</table>
|
|
|
103 |
<tr>
|
104 |
<th class="tg-0pky">APK</th>
|
105 |
<th class="tg-0lax">Comment</th>
|
106 |
+
<th class="tg-0pky">Model</th>
|
|
|
107 |
</tr>
|
108 |
</thead>
|
109 |
<tbody>
|
110 |
<tr>
|
111 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-bilingual_zh_en-zipformer.apk</td>
|
112 |
<td class="tg-0lax">It supports both English and Chinese.</td>
|
113 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2">sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2</a></td>
|
|
|
114 |
</tr>
|
115 |
<tr>
|
116 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-fr-zipformer.apk</td>
|
117 |
<td class="tg-0lax"><span style="font-weight:400;font-style:normal">It supports only French.</span></td>
|
118 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-fr-2023-04-14.tar.bz2">sherpa-onnx-streaming-zipformer-fr-2023-04-14.tar.bz2</a></td>
|
|
|
119 |
</tr>
|
120 |
<tr>
|
121 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-zh-zipformer2.apk</td>
|
122 |
<td class="tg-0lax">It supports only Chinese.</td>
|
123 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/icefall-asr-zipformer-streaming-wenetspeech-20230615.tar.bz2">icefall-asr-zipformer-streaming-wenetspeech-20230615.tar.bz2</a></td>
|
|
|
124 |
</tr>
|
125 |
<tr>
|
126 |
+
<td class="tg-0pky">sherpa-onnx-x.y.zz-arm64-v8a-asr-zh-small_zipformer.apk</td>
|
127 |
<td class="tg-0lax">It supports only Chinese.<br>It uses the smallest zipformer and runs<br>super fast, though its accuracy is not that good.</td>
|
128 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-zh-14M-2023-02-23.tar.bz2">sherpa-onnx-streaming-zipformer-zh-14M-2023-02-23.tar.bz2</a></td>
|
|
|
129 |
</tr>
|
130 |
<tr>
|
131 |
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-asr-en-small_zipformer.apk</td>
|
132 |
<td class="tg-0lax">It suppors only English.<br>It uses a very small zipformer and runs<br>super fast, though its accuracy is not that good.</td>
|
133 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-en-20M-2023-02-17.tar.bz2">sherpa-onnx-streaming-zipformer-en-20M-2023-02-17.tar.bz2</a></td>
|
|
|
134 |
</tr>
|
135 |
</tbody>
|
136 |
</table>
|
generate-kws.py
ADDED
@@ -0,0 +1,158 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
import os
|
3 |
+
import re
|
4 |
+
from pathlib import Path
|
5 |
+
from typing import List
|
6 |
+
|
7 |
+
BASE_URL = "https://huggingface.co/csukuangfj/sherpa-onnx-apk/resolve/main/"
|
8 |
+
|
9 |
+
from dataclasses import dataclass
|
10 |
+
|
11 |
+
|
12 |
+
@dataclass
|
13 |
+
class APK:
|
14 |
+
major: int
|
15 |
+
minor: int
|
16 |
+
patch: int
|
17 |
+
arch: str
|
18 |
+
short_name: str
|
19 |
+
|
20 |
+
def __init__(self, s):
|
21 |
+
# sherpa-onnx-1.9.23-arm64-v8a-kws-en-gigaspeech-zipformer.apk
|
22 |
+
# sherpa-onnx-1.9.23-x86-kws-en-gigaspeech-zipformer.apk
|
23 |
+
s = str(s)[len("asr/") :]
|
24 |
+
split = s.split("-")
|
25 |
+
self.major, self.minor, self.patch = list(map(int, split[2].split(".")))
|
26 |
+
self.arch = split[3]
|
27 |
+
self.lang = split[5]
|
28 |
+
self.short_name = split[6] + split[7]
|
29 |
+
if "arm" in s:
|
30 |
+
self.arch += "-" + split[4]
|
31 |
+
self.lang = split[6]
|
32 |
+
self.short_name = split[7] + split[8]
|
33 |
+
|
34 |
+
if "armeabi" in self.arch:
|
35 |
+
self.arch = "y" + self.arch
|
36 |
+
|
37 |
+
if "arm64" in self.arch:
|
38 |
+
self.arch = "z" + self.arch
|
39 |
+
|
40 |
+
|
41 |
+
def sort_by_apk(x):
|
42 |
+
x = APK(x)
|
43 |
+
return (x.major, x.minor, x.patch, x.arch, x.lang, x.short_name)
|
44 |
+
|
45 |
+
|
46 |
+
def generate_url(files: List[str]) -> List[str]:
|
47 |
+
ans = []
|
48 |
+
base = BASE_URL
|
49 |
+
for f in files:
|
50 |
+
ans.append(base + str(f))
|
51 |
+
return ans
|
52 |
+
|
53 |
+
|
54 |
+
def get_all_files(d: str, suffix: str) -> List[str]:
|
55 |
+
ans = sorted(Path(d).glob(suffix), key=sort_by_apk, reverse=True)
|
56 |
+
return list(map(lambda x: BASE_URL + str(x), ans))
|
57 |
+
|
58 |
+
|
59 |
+
def to_file(filename: str, files: List[str]):
|
60 |
+
content = r"""
|
61 |
+
<h1> APKs for keyword spotting </h1>
|
62 |
+
This page lists the <strong>keyword spotting</strong> APKs for <a href="http://github.com/k2-fsa/sherpa-onnx">sherpa-onnx</a>,
|
63 |
+
one of the deployment frameworks of <a href="https://github.com/k2-fsa">the Next-gen Kaldi project</a>.
|
64 |
+
<br/>
|
65 |
+
The name of an APK has the following rule:
|
66 |
+
<ul>
|
67 |
+
<li> sherpa-onnx-{version}-{arch}-kws-${lang}-{dataset}-{model}.apk
|
68 |
+
</ul>
|
69 |
+
where
|
70 |
+
<ul>
|
71 |
+
<li> version: It specifies the current version, e.g., 1.9.23
|
72 |
+
<li> arch: The architecture targeted by this APK, e.g., arm64-v8a, armeabi-v7a, x86_64, x86
|
73 |
+
<li> lang: The lang of the model used in the APK, e.g., en for English, zh for Chinese
|
74 |
+
<li> dataset: The dataset used to train the model
|
75 |
+
<li> model: The name of the model used in the APK
|
76 |
+
</ul>
|
77 |
+
|
78 |
+
<br/>
|
79 |
+
|
80 |
+
You can download all supported models from
|
81 |
+
<a href="https://github.com/k2-fsa/sherpa-onnx/releases/tag/kws-models">https://github.com/k2-fsa/sherpa-onnx/releases/tag/kws-models</a>
|
82 |
+
|
83 |
+
<br/>
|
84 |
+
<br/>
|
85 |
+
|
86 |
+
<!--
|
87 |
+
see https://www.tablesgenerator.com/html_tables#
|
88 |
+
-->
|
89 |
+
|
90 |
+
<style type="text/css">
|
91 |
+
.tg {border-collapse:collapse;border-spacing:0;}
|
92 |
+
.tg td{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
|
93 |
+
overflow:hidden;padding:10px 5px;word-break:normal;}
|
94 |
+
.tg th{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
|
95 |
+
font-weight:normal;overflow:hidden;padding:10px 5px;word-break:normal;}
|
96 |
+
.tg .tg-0pky{border-color:inherit;text-align:left;vertical-align:top}
|
97 |
+
.tg .tg-0lax{text-align:left;vertical-align:top}
|
98 |
+
</style>
|
99 |
+
<table class="tg">
|
100 |
+
<thead>
|
101 |
+
<tr>
|
102 |
+
<th class="tg-0pky">APK</th>
|
103 |
+
<th class="tg-0lax">Comment</th>
|
104 |
+
<th class="tg-0pky">Model</th>
|
105 |
+
</tr>
|
106 |
+
</thead>
|
107 |
+
<tbody>
|
108 |
+
<tr>
|
109 |
+
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-kws-zh-wenetspeech-zipformer.apk</td>
|
110 |
+
<td class="tg-0lax">It supports only Chinese.</td>
|
111 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/kws-models/sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01.tar.bz2">sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01.tar.bz2</a></td>
|
112 |
+
</tr>
|
113 |
+
<tr>
|
114 |
+
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-kws-en-gigaspeech-zipformer.apk</td>
|
115 |
+
<td class="tg-0lax">It supports only English.</td>
|
116 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/kws-models/sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2">sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2</a></td>
|
117 |
+
</tr>
|
118 |
+
</tbody>
|
119 |
+
</table>
|
120 |
+
|
121 |
+
<br/>
|
122 |
+
<br/>
|
123 |
+
|
124 |
+
<div/>
|
125 |
+
"""
|
126 |
+
if "-cn" not in filename:
|
127 |
+
content += """
|
128 |
+
For Chinese users, please <a href="./apk-cn.html">visit this address</a>,
|
129 |
+
which replaces <a href="huggingface.co">huggingface.co</a> with <a href="hf-mirror.com">hf-mirror.com</a>
|
130 |
+
<br/>
|
131 |
+
<br/>
|
132 |
+
中国用户, 请访问<a href="./apk-cn.html">这个地址</a>
|
133 |
+
<br/>
|
134 |
+
<br/>
|
135 |
+
"""
|
136 |
+
|
137 |
+
with open(filename, "w") as f:
|
138 |
+
print(content, file=f)
|
139 |
+
for x in files:
|
140 |
+
name = x.rsplit("/", maxsplit=1)[-1]
|
141 |
+
print(f'<a href="{x}" />{name}<br/>', file=f)
|
142 |
+
|
143 |
+
|
144 |
+
def main():
|
145 |
+
apk = get_all_files("kws", suffix="*.apk")
|
146 |
+
to_file("./apk-kws.html", apk)
|
147 |
+
|
148 |
+
# for Chinese users
|
149 |
+
apk2 = []
|
150 |
+
for a in apk:
|
151 |
+
a = a.replace("huggingface.co", "hf-mirror.com")
|
152 |
+
apk2.append(a)
|
153 |
+
|
154 |
+
to_file("./apk-kws-cn.html", apk2)
|
155 |
+
|
156 |
+
|
157 |
+
if __name__ == "__main__":
|
158 |
+
main()
|
generate-vad-asr.py
ADDED
@@ -0,0 +1,169 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
import os
|
3 |
+
import re
|
4 |
+
from pathlib import Path
|
5 |
+
from typing import List
|
6 |
+
|
7 |
+
BASE_URL = "https://huggingface.co/csukuangfj/sherpa-onnx-apk/resolve/main/"
|
8 |
+
|
9 |
+
from dataclasses import dataclass
|
10 |
+
|
11 |
+
|
12 |
+
@dataclass
|
13 |
+
class APK:
|
14 |
+
major: int
|
15 |
+
minor: int
|
16 |
+
patch: int
|
17 |
+
arch: str
|
18 |
+
short_name: str
|
19 |
+
|
20 |
+
def __init__(self, s):
|
21 |
+
# sherpa-onnx-1.9.23-arm64-v8a-vad_asr-en-whisper_tiny.apk
|
22 |
+
# sherpa-onnx-1.9.23-x86-vad_asr-en-whisper_tiny.apk
|
23 |
+
s = str(s)[len("vad-asr/") :]
|
24 |
+
split = s.split("-")
|
25 |
+
self.major, self.minor, self.patch = list(map(int, split[2].split(".")))
|
26 |
+
self.arch = split[3]
|
27 |
+
self.lang = split[5]
|
28 |
+
self.short_name = split[6]
|
29 |
+
if "arm" in s:
|
30 |
+
self.arch += "-" + split[4]
|
31 |
+
self.lang = split[6]
|
32 |
+
self.short_name = split[7]
|
33 |
+
|
34 |
+
if "armeabi" in self.arch:
|
35 |
+
self.arch = "y" + self.arch
|
36 |
+
|
37 |
+
if "arm64" in self.arch:
|
38 |
+
self.arch = "z" + self.arch
|
39 |
+
|
40 |
+
if "small" in self.short_name:
|
41 |
+
self.short_name = "zzz" + self.short_name
|
42 |
+
|
43 |
+
|
44 |
+
def sort_by_apk(x):
|
45 |
+
x = APK(x)
|
46 |
+
return (x.major, x.minor, x.patch, x.arch, x.lang, x.short_name)
|
47 |
+
|
48 |
+
|
49 |
+
def generate_url(files: List[str]) -> List[str]:
|
50 |
+
ans = []
|
51 |
+
base = BASE_URL
|
52 |
+
for f in files:
|
53 |
+
ans.append(base + str(f))
|
54 |
+
return ans
|
55 |
+
|
56 |
+
|
57 |
+
def get_all_files(d: str, suffix: str) -> List[str]:
|
58 |
+
ans = sorted(Path(d).glob(suffix), key=sort_by_apk, reverse=True)
|
59 |
+
return list(map(lambda x: BASE_URL + str(x), ans))
|
60 |
+
|
61 |
+
|
62 |
+
def to_file(filename: str, files: List[str]):
|
63 |
+
content = r"""
|
64 |
+
<h1> APKs for VAD + non-streaming speech recognition </h1>
|
65 |
+
This page lists the <strong>VAD + non-streaming speech recognition</strong> APKs for <a href="http://github.com/k2-fsa/sherpa-onnx">sherpa-onnx</a>,
|
66 |
+
one of the deployment frameworks of <a href="https://github.com/k2-fsa">the Next-gen Kaldi project</a>.
|
67 |
+
<br/>
|
68 |
+
The name of an APK has the following rule:
|
69 |
+
<ul>
|
70 |
+
<li> sherpa-onnx-{version}-{arch}-vad_asr-{lang}-{model}.apk
|
71 |
+
</ul>
|
72 |
+
where
|
73 |
+
<ul>
|
74 |
+
<li> version: It specifies the current version, e.g., 1.9.23
|
75 |
+
<li> arch: The architecture targeted by this APK, e.g., arm64-v8a, armeabi-v7a, x86_64, x86
|
76 |
+
<li> lang: The lang of the model used in the APK, e.g., en for English, zh for Chinese
|
77 |
+
<li> model: The name of the model used in the APK
|
78 |
+
</ul>
|
79 |
+
|
80 |
+
<br/>
|
81 |
+
|
82 |
+
You can download all supported models from
|
83 |
+
<a href="https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models">https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models</a>
|
84 |
+
|
85 |
+
<br/>
|
86 |
+
<br/>
|
87 |
+
|
88 |
+
<!--
|
89 |
+
see https://www.tablesgenerator.com/html_tables#
|
90 |
+
-->
|
91 |
+
|
92 |
+
<style type="text/css">
|
93 |
+
.tg {border-collapse:collapse;border-spacing:0;}
|
94 |
+
.tg td{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
|
95 |
+
overflow:hidden;padding:10px 5px;word-break:normal;}
|
96 |
+
.tg th{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
|
97 |
+
font-weight:normal;overflow:hidden;padding:10px 5px;word-break:normal;}
|
98 |
+
.tg .tg-0pky{border-color:inherit;text-align:left;vertical-align:top}
|
99 |
+
.tg .tg-0lax{text-align:left;vertical-align:top}
|
100 |
+
</style>
|
101 |
+
<table class="tg">
|
102 |
+
<thead>
|
103 |
+
<tr>
|
104 |
+
<th class="tg-0pky">APK</th>
|
105 |
+
<th class="tg-0lax">Comment</th>
|
106 |
+
<th class="tg-0pky">VAD model</th>
|
107 |
+
<th class="tg-0pky">Non-streaming ASR model</th>
|
108 |
+
</tr>
|
109 |
+
</thead>
|
110 |
+
<tbody>
|
111 |
+
<tr>
|
112 |
+
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-vad_asr-zh-zipformer.apk</td>
|
113 |
+
<td class="tg-0lax">It supports only Chinese.</td>
|
114 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx">silero_vad.onnx</a></td>
|
115 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/icefall-asr-zipformer-wenetspeech-20230615.tar.bz2">icefall-asr-zipformer-wenetspeech-20230615</a></td>
|
116 |
+
</tr>
|
117 |
+
<tr>
|
118 |
+
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-vad_asr-zh-paraformer.apk</td>
|
119 |
+
<td class="tg-0lax"><span style="font-weight:400;font-style:normal">It supports both Chinese and English.</span></td>
|
120 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx">silero_vad.onnx</a></td>
|
121 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2">sherpa-onnx-paraformer-zh-2023-03-28</a></td>
|
122 |
+
</tr>
|
123 |
+
<tr>
|
124 |
+
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-vad_asr-en-whisper_tiny.apk</td>
|
125 |
+
<td class="tg-0lax">It supports only English.</td>
|
126 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx">silero_vad.onnx</a></td>
|
127 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-whisper-tiny.en.tar.bz2">sherpa-onnx-whisper-tiny.en</a></td>
|
128 |
+
</tr>
|
129 |
+
</tbody>
|
130 |
+
</table>
|
131 |
+
|
132 |
+
<br/>
|
133 |
+
<br/>
|
134 |
+
|
135 |
+
<div/>
|
136 |
+
"""
|
137 |
+
if "-cn" not in filename:
|
138 |
+
content += """
|
139 |
+
For Chinese users, please <a href="./apk-vad-asr-cn.html">visit this address</a>,
|
140 |
+
which replaces <a href="huggingface.co">huggingface.co</a> with <a href="hf-mirror.com">hf-mirror.com</a>
|
141 |
+
<br/>
|
142 |
+
<br/>
|
143 |
+
中国用户, 请访问<a href="./apk-vad-asr.cn.html">这个地址</a>
|
144 |
+
<br/>
|
145 |
+
<br/>
|
146 |
+
"""
|
147 |
+
|
148 |
+
with open(filename, "w") as f:
|
149 |
+
print(content, file=f)
|
150 |
+
for x in files:
|
151 |
+
name = x.rsplit("/", maxsplit=1)[-1]
|
152 |
+
print(f'<a href="{x}" />{name}<br/>', file=f)
|
153 |
+
|
154 |
+
|
155 |
+
def main():
|
156 |
+
apk = get_all_files("vad-asr", suffix="*.apk")
|
157 |
+
to_file("./apk-vad-asr.html", apk)
|
158 |
+
|
159 |
+
# for Chinese users
|
160 |
+
apk2 = []
|
161 |
+
for a in apk:
|
162 |
+
a = a.replace("huggingface.co", "hf-mirror.com")
|
163 |
+
apk2.append(a)
|
164 |
+
|
165 |
+
to_file("./apk-vad-asr-cn.html", apk2)
|
166 |
+
|
167 |
+
|
168 |
+
if __name__ == "__main__":
|
169 |
+
main()
|
generate-vad.py
ADDED
@@ -0,0 +1,149 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
import os
|
3 |
+
import re
|
4 |
+
from pathlib import Path
|
5 |
+
from typing import List
|
6 |
+
|
7 |
+
BASE_URL = "https://huggingface.co/csukuangfj/sherpa-onnx-apk/resolve/main/"
|
8 |
+
|
9 |
+
from dataclasses import dataclass
|
10 |
+
|
11 |
+
|
12 |
+
@dataclass
|
13 |
+
class APK:
|
14 |
+
major: int
|
15 |
+
minor: int
|
16 |
+
patch: int
|
17 |
+
arch: str
|
18 |
+
short_name: str
|
19 |
+
|
20 |
+
def __init__(self, s):
|
21 |
+
# sherpa-onnx-1.9.23-arm64-v8a-silero_vad.apk
|
22 |
+
# sherpa-onnx-1.9.23-x86-silero_vad.apk
|
23 |
+
s = str(s)[len("asr/") :]
|
24 |
+
split = s.split("-")
|
25 |
+
self.major, self.minor, self.patch = list(map(int, split[2].split(".")))
|
26 |
+
self.arch = split[3]
|
27 |
+
self.short_name = split[4]
|
28 |
+
if "arm" in s:
|
29 |
+
self.arch += "-" + split[4]
|
30 |
+
self.short_name = split[5]
|
31 |
+
|
32 |
+
if "armeabi" in self.arch:
|
33 |
+
self.arch = "y" + self.arch
|
34 |
+
|
35 |
+
if "arm64" in self.arch:
|
36 |
+
self.arch = "z" + self.arch
|
37 |
+
|
38 |
+
|
39 |
+
def sort_by_apk(x):
|
40 |
+
x = APK(x)
|
41 |
+
return (x.major, x.minor, x.patch, x.arch, x.short_name)
|
42 |
+
|
43 |
+
|
44 |
+
def generate_url(files: List[str]) -> List[str]:
|
45 |
+
ans = []
|
46 |
+
base = BASE_URL
|
47 |
+
for f in files:
|
48 |
+
ans.append(base + str(f))
|
49 |
+
return ans
|
50 |
+
|
51 |
+
|
52 |
+
def get_all_files(d: str, suffix: str) -> List[str]:
|
53 |
+
ans = sorted(Path(d).glob(suffix), key=sort_by_apk, reverse=True)
|
54 |
+
return list(map(lambda x: BASE_URL + str(x), ans))
|
55 |
+
|
56 |
+
|
57 |
+
def to_file(filename: str, files: List[str]):
|
58 |
+
content = r"""
|
59 |
+
<h1> APKs for VAD </h1>
|
60 |
+
This page lists the <strong>VAD</strong> APKs for <a href="http://github.com/k2-fsa/sherpa-onnx">sherpa-onnx</a>,
|
61 |
+
one of the deployment frameworks of <a href="https://github.com/k2-fsa">the Next-gen Kaldi project</a>.
|
62 |
+
<br/>
|
63 |
+
The name of an APK has the following rule:
|
64 |
+
<ul>
|
65 |
+
<li> sherpa-onnx-{version}-{arch}-{model}.apk
|
66 |
+
</ul>
|
67 |
+
where
|
68 |
+
<ul>
|
69 |
+
<li> version: It specifies the current version, e.g., 1.9.23
|
70 |
+
<li> arch: The architecture targeted by this APK, e.g., arm64-v8a, armeabi-v7a, x86_64, x86
|
71 |
+
<li> model: The name of the model used in the APK
|
72 |
+
</ul>
|
73 |
+
|
74 |
+
<br/>
|
75 |
+
|
76 |
+
You can download all supported models from
|
77 |
+
<a href="https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models">https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models</a>
|
78 |
+
|
79 |
+
<br/>
|
80 |
+
<br/>
|
81 |
+
|
82 |
+
<!--
|
83 |
+
see https://www.tablesgenerator.com/html_tables#
|
84 |
+
-->
|
85 |
+
|
86 |
+
<style type="text/css">
|
87 |
+
.tg {border-collapse:collapse;border-spacing:0;}
|
88 |
+
.tg td{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
|
89 |
+
overflow:hidden;padding:10px 5px;word-break:normal;}
|
90 |
+
.tg th{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px;
|
91 |
+
font-weight:normal;overflow:hidden;padding:10px 5px;word-break:normal;}
|
92 |
+
.tg .tg-0pky{border-color:inherit;text-align:left;vertical-align:top}
|
93 |
+
.tg .tg-0lax{text-align:left;vertical-align:top}
|
94 |
+
</style>
|
95 |
+
<table class="tg">
|
96 |
+
<thead>
|
97 |
+
<tr>
|
98 |
+
<th class="tg-0pky">APK</th>
|
99 |
+
<th class="tg-0lax">Comment</th>
|
100 |
+
<th class="tg-0pky">Model</th>
|
101 |
+
</tr>
|
102 |
+
</thead>
|
103 |
+
<tbody>
|
104 |
+
<tr>
|
105 |
+
<td class="tg-0pky">sherpa-onnx-x.y.z-arm64-v8a-silero_vad.apk</td>
|
106 |
+
<td class="tg-0lax">It uses <a href="https://github.com/snakers4/silero-vad">https://github.com/snakers4/silero-vad</a></td>
|
107 |
+
<td class="tg-0pky"><a href="https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx">silero_vad.onnx</a></td>
|
108 |
+
</tr>
|
109 |
+
</tbody>
|
110 |
+
</table>
|
111 |
+
|
112 |
+
<br/>
|
113 |
+
<br/>
|
114 |
+
|
115 |
+
<div/>
|
116 |
+
"""
|
117 |
+
if "-cn" not in filename:
|
118 |
+
content += """
|
119 |
+
For Chinese users, please <a href="./apk-cn.html">visit this address</a>,
|
120 |
+
which replaces <a href="huggingface.co">huggingface.co</a> with <a href="hf-mirror.com">hf-mirror.com</a>
|
121 |
+
<br/>
|
122 |
+
<br/>
|
123 |
+
中国用户, 请访问<a href="./apk-cn.html">这个地址</a>
|
124 |
+
<br/>
|
125 |
+
<br/>
|
126 |
+
"""
|
127 |
+
|
128 |
+
with open(filename, "w") as f:
|
129 |
+
print(content, file=f)
|
130 |
+
for x in files:
|
131 |
+
name = x.rsplit("/", maxsplit=1)[-1]
|
132 |
+
print(f'<a href="{x}" />{name}<br/>', file=f)
|
133 |
+
|
134 |
+
|
135 |
+
def main():
|
136 |
+
apk = get_all_files("vad", suffix="*.apk")
|
137 |
+
to_file("./apk-vad.html", apk)
|
138 |
+
|
139 |
+
# for Chinese users
|
140 |
+
apk2 = []
|
141 |
+
for a in apk:
|
142 |
+
a = a.replace("huggingface.co", "hf-mirror.com")
|
143 |
+
apk2.append(a)
|
144 |
+
|
145 |
+
to_file("./apk-vad-cn.html", apk2)
|
146 |
+
|
147 |
+
|
148 |
+
if __name__ == "__main__":
|
149 |
+
main()
|
kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-arm64-v8a.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:724dcf9b581591c4f66e89c4e38171eab260669f07ce007937872d06ab0526b0
|
3 |
-
size 26877670
|
|
|
|
|
|
|
|
kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-armeabi-v7a.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:c37822c1bb4d6ac90110697ccdbe70ebe32462bd50e2bca76abd1df6026e31b8
|
3 |
-
size 25526012
|
|
|
|
|
|
|
|
kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-x86.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:6e6b8d9abf62e4f629439ebba97991ccfbd31ee5fc0be379dc2525543f5ade79
|
3 |
-
size 28098212
|
|
|
|
|
|
|
|
kws/sherpa-onnx-kws-gigaspeech-en-1.9.23-x86_64.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:0cadd76e29d0223ec2aaad5edf4d67732e7edc2cd60685789c302e88fe78b7b6
|
3 |
-
size 27803333
|
|
|
|
|
|
|
|
kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-arm64-v8a.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:b193096bdc0fdd8da51e05d6125a266fd688ff59a02d3d847883239e4c8ce277
|
3 |
-
size 26054265
|
|
|
|
|
|
|
|
kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-armeabi-v7a.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:3acf03d502b23dd4b7982f1fcc26695c332b1a9de5de24b30dd8f62e5e7cbd1b
|
3 |
-
size 24706703
|
|
|
|
|
|
|
|
kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-x86.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:6ef5a6c868ed9bfbbf829b98890e912d271a56cca9b23697640947952bf322a4
|
3 |
-
size 27278903
|
|
|
|
|
|
|
|
kws/sherpa-onnx-kws-wenetspeech-zh-1.9.23-x86_64.apk
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:0f2faa7df7dac2486833e605abc561657030cd1633f197b8eacf5429bf17aad3
|
3 |
-
size 26979928
|
|
|
|
|
|
|
|