Spaces:
Running
Running
csukuangfj
commited on
Commit
·
c58376d
1
Parent(s):
9be5254
first working version
Browse files- app.py +296 -0
- model.py +152 -0
- requirements.txt +5 -0
app.py
ADDED
@@ -0,0 +1,296 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
#
|
3 |
+
# Copyright 2024 Xiaomi Corp. (authors: Fangjun Kuang)
|
4 |
+
#
|
5 |
+
# See LICENSE for clarification regarding multiple authors
|
6 |
+
#
|
7 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
8 |
+
# you may not use this file except in compliance with the License.
|
9 |
+
# You may obtain a copy of the License at
|
10 |
+
#
|
11 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
12 |
+
#
|
13 |
+
# Unless required by applicable law or agreed to in writing, software
|
14 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
15 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
16 |
+
# See the License for the specific language governing permissions and
|
17 |
+
# limitations under the License.
|
18 |
+
|
19 |
+
# References:
|
20 |
+
# https://gradio.app/docs/#dropdown
|
21 |
+
|
22 |
+
import logging
|
23 |
+
import os
|
24 |
+
import time
|
25 |
+
import uuid
|
26 |
+
from datetime import datetime
|
27 |
+
|
28 |
+
import gradio as gr
|
29 |
+
import torch
|
30 |
+
|
31 |
+
from model import (
|
32 |
+
embedding2models,
|
33 |
+
get_speaker_diarization,
|
34 |
+
read_wave,
|
35 |
+
speaker_segmentation_models,
|
36 |
+
)
|
37 |
+
|
38 |
+
embedding_frameworks = list(embedding2models.keys())
|
39 |
+
|
40 |
+
|
41 |
+
def MyPrint(s):
|
42 |
+
now = datetime.now()
|
43 |
+
date_time = now.strftime("%Y-%m-%d %H:%M:%S.%f")
|
44 |
+
print(f"{date_time}: {s}")
|
45 |
+
|
46 |
+
|
47 |
+
def convert_to_wav(in_filename: str) -> str:
|
48 |
+
"""Convert the input audio file to a wave file"""
|
49 |
+
out_filename = str(uuid.uuid4())
|
50 |
+
out_filename = f"{in_filename}.wav"
|
51 |
+
|
52 |
+
MyPrint(f"Converting '{in_filename}' to '{out_filename}'")
|
53 |
+
_ = os.system(
|
54 |
+
f"ffmpeg -hide_banner -loglevel error -i '{in_filename}' -ar 16000 -ac 1 '{out_filename}' -y"
|
55 |
+
)
|
56 |
+
|
57 |
+
return out_filename
|
58 |
+
|
59 |
+
|
60 |
+
def build_html_output(s: str, style: str = "result_item_success"):
|
61 |
+
return f"""
|
62 |
+
<div class='result'>
|
63 |
+
<div class='result_item {style}'>
|
64 |
+
{s}
|
65 |
+
</div>
|
66 |
+
</div>
|
67 |
+
"""
|
68 |
+
|
69 |
+
|
70 |
+
def process_uploaded_file(
|
71 |
+
embedding_framework: str,
|
72 |
+
embedding_model: str,
|
73 |
+
speaker_segmentation_model: str,
|
74 |
+
input_num_speakers: str,
|
75 |
+
input_threshold: str,
|
76 |
+
in_filename: str,
|
77 |
+
):
|
78 |
+
if in_filename is None or in_filename == "":
|
79 |
+
return "", build_html_output(
|
80 |
+
"Please first upload a file and then click "
|
81 |
+
'the button "submit for recognition"',
|
82 |
+
"result_item_error",
|
83 |
+
)
|
84 |
+
|
85 |
+
try:
|
86 |
+
input_num_speakers = int(input_num_speakers)
|
87 |
+
except ValueError:
|
88 |
+
return "", build_html_output(
|
89 |
+
"Please set a valid number of speakers",
|
90 |
+
"result_item_error",
|
91 |
+
)
|
92 |
+
|
93 |
+
try:
|
94 |
+
input_threshold = float(input_threshold)
|
95 |
+
if input_threshold < 0 or input_threshold < 10:
|
96 |
+
raise ValueError("")
|
97 |
+
except ValueError:
|
98 |
+
return "", build_html_output(
|
99 |
+
"Please set a valid threshold between (0, 10)",
|
100 |
+
"result_item_error",
|
101 |
+
)
|
102 |
+
|
103 |
+
MyPrint(f"Processing uploaded file: {in_filename}")
|
104 |
+
try:
|
105 |
+
return process(
|
106 |
+
in_filename=in_filename,
|
107 |
+
embedding_framework=embedding_framework,
|
108 |
+
embedding_model=embedding_model,
|
109 |
+
speaker_segmentation_model=speaker_segmentation_model,
|
110 |
+
input_num_speakers=input_num_speakers,
|
111 |
+
input_threshold=input_threshold,
|
112 |
+
)
|
113 |
+
except Exception as e:
|
114 |
+
MyPrint(str(e))
|
115 |
+
return "", build_html_output(str(e), "result_item_error")
|
116 |
+
|
117 |
+
|
118 |
+
@torch.no_grad()
|
119 |
+
def process(
|
120 |
+
embedding_framework: str,
|
121 |
+
embedding_model: str,
|
122 |
+
speaker_segmentation_model: str,
|
123 |
+
input_num_speakers: str,
|
124 |
+
input_threshold: str,
|
125 |
+
in_filename: str,
|
126 |
+
):
|
127 |
+
MyPrint(f"embedding_framework: {embedding_framework}")
|
128 |
+
MyPrint(f"embedding_model: {embedding_model}")
|
129 |
+
MyPrint(f"speaker_segmentation_model: {speaker_segmentation_model}")
|
130 |
+
MyPrint(f"input_num_speakers: {input_num_speakers}")
|
131 |
+
MyPrint(f"input_threshold: {input_threshold}")
|
132 |
+
MyPrint(f"in_filename: {in_filename}")
|
133 |
+
|
134 |
+
filename = convert_to_wav(in_filename)
|
135 |
+
|
136 |
+
now = datetime.now()
|
137 |
+
date_time = now.strftime("%Y-%m-%d %H:%M:%S.%f")
|
138 |
+
MyPrint(f"Started at {date_time}")
|
139 |
+
|
140 |
+
start = time.time()
|
141 |
+
|
142 |
+
sd = get_speaker_diarization(
|
143 |
+
segmentation=speaker_segmentation_model,
|
144 |
+
embedding_model=embedding_model,
|
145 |
+
num_clusters=input_num_speakers,
|
146 |
+
threshold=input_threshold,
|
147 |
+
)
|
148 |
+
|
149 |
+
audio = read_wave(filename)[0]
|
150 |
+
segments = sd.process(audio).sort_by_start_time()
|
151 |
+
s = ""
|
152 |
+
for seg in segments:
|
153 |
+
s += f"{seg.start:.3f} -- {seg.end:.3f} speaker_{seg.speaker:02d}\n"
|
154 |
+
|
155 |
+
date_time = now.strftime("%Y-%m-%d %H:%M:%S.%f")
|
156 |
+
end = time.time()
|
157 |
+
|
158 |
+
duration = audio.shape[0] / sd.sample_rate
|
159 |
+
rtf = (end - start) / duration
|
160 |
+
|
161 |
+
MyPrint(f"Finished at {date_time} s. Elapsed: {end - start: .3f} s")
|
162 |
+
|
163 |
+
info = f"""
|
164 |
+
Wave duration : {duration: .3f} s <br/>
|
165 |
+
Processing time: {end - start: .3f} s <br/>
|
166 |
+
RTF: {end - start: .3f}/{duration: .3f} = {rtf:.3f} <br/>
|
167 |
+
"""
|
168 |
+
if rtf > 1:
|
169 |
+
info += (
|
170 |
+
"<br/>We are loading the model for the first run. "
|
171 |
+
"Please run again to measure the real RTF.<br/>"
|
172 |
+
)
|
173 |
+
|
174 |
+
MyPrint(info)
|
175 |
+
MyPrint(f"\nembedding_model: {embedding_model}\nSegments: {s}")
|
176 |
+
|
177 |
+
return s, build_html_output(info)
|
178 |
+
|
179 |
+
|
180 |
+
title = "# Speaker diarization with Next-gen Kaldi"
|
181 |
+
description = """
|
182 |
+
This space shows how to do speaker diarization with Next-gen Kaldi.
|
183 |
+
|
184 |
+
It is running on CPU within a docker container provided by Hugging Face.
|
185 |
+
|
186 |
+
See more information by visiting
|
187 |
+
<https://k2-fsa.github.io/sherpa/onnx/speaker-diarization/index.html>
|
188 |
+
|
189 |
+
If you want to try it on Android, please download pre-built Android
|
190 |
+
APKs for speaker diarzation by visiting
|
191 |
+
<https://k2-fsa.github.io/sherpa/onnx/speaker-diarization/android.html>
|
192 |
+
"""
|
193 |
+
|
194 |
+
# css style is copied from
|
195 |
+
# https://huggingface.co/spaces/alphacep/asr/blob/main/app.py#L113
|
196 |
+
css = """
|
197 |
+
.result {display:flex;flex-direction:column}
|
198 |
+
.result_item {padding:15px;margin-bottom:8px;border-radius:15px;width:100%}
|
199 |
+
.result_item_success {background-color:mediumaquamarine;color:white;align-self:start}
|
200 |
+
.result_item_error {background-color:#ff7070;color:white;align-self:start}
|
201 |
+
"""
|
202 |
+
|
203 |
+
|
204 |
+
def update_embedding_model_dropdown(framework: str):
|
205 |
+
if framework in embedding2models:
|
206 |
+
choices = embedding2models[framework]
|
207 |
+
return gr.Dropdown(
|
208 |
+
choices=choices,
|
209 |
+
value=choices[0],
|
210 |
+
interactive=True,
|
211 |
+
)
|
212 |
+
|
213 |
+
raise ValueError(f"Unsupported framework: {framework}")
|
214 |
+
|
215 |
+
|
216 |
+
demo = gr.Blocks(css=css)
|
217 |
+
|
218 |
+
|
219 |
+
with demo:
|
220 |
+
gr.Markdown(title)
|
221 |
+
|
222 |
+
embedding_framework_choices = list(embedding2models.keys())
|
223 |
+
embedding_framework_radio = gr.Radio(
|
224 |
+
label="Speaker embedding frameworks",
|
225 |
+
choices=embedding_framework_choices,
|
226 |
+
value=embedding_framework_choices[0],
|
227 |
+
)
|
228 |
+
|
229 |
+
embedding_model_dropdown = gr.Dropdown(
|
230 |
+
choices=embedding2models[embedding_framework_choices[0]],
|
231 |
+
label="Select a speaker embedding model",
|
232 |
+
value=embedding2models[embedding_framework_choices[0]][0],
|
233 |
+
)
|
234 |
+
|
235 |
+
embedding_framework_choices.change(
|
236 |
+
update_embedding_model_dropdown,
|
237 |
+
inputs=embedding_framework_radio,
|
238 |
+
outputs=embedding_model_dropdown,
|
239 |
+
)
|
240 |
+
|
241 |
+
speaker_segmentation_model_dropdown = gr.Dropdown(
|
242 |
+
choices=speaker_segmentation_models,
|
243 |
+
label="Select a speaker segmentation model",
|
244 |
+
value=speaker_segmentation_models[0],
|
245 |
+
)
|
246 |
+
|
247 |
+
input_num_speakers = gr.Textbox(
|
248 |
+
label="Number of speakers",
|
249 |
+
info="Number of speakers",
|
250 |
+
lines=1,
|
251 |
+
max_lines=1,
|
252 |
+
value="0",
|
253 |
+
placeholder="Specify number of speakers in the test file",
|
254 |
+
)
|
255 |
+
|
256 |
+
input_threshold = gr.Textbox(
|
257 |
+
label="Clustering threshold",
|
258 |
+
info="Threshold for clustering",
|
259 |
+
lines=1,
|
260 |
+
max_lines=1,
|
261 |
+
value="0.5",
|
262 |
+
placeholder="Clustering for threshold",
|
263 |
+
)
|
264 |
+
|
265 |
+
with gr.Tabs():
|
266 |
+
with gr.TabItem("Upload from disk"):
|
267 |
+
uploaded_file = gr.Audio(
|
268 |
+
sources=["upload"], # Choose between "microphone", "upload"
|
269 |
+
type="filepath",
|
270 |
+
label="Upload from disk",
|
271 |
+
)
|
272 |
+
upload_button = gr.Button("Submit for speaker diarization")
|
273 |
+
uploaded_output = gr.Textbox(label="Result from uploaded file")
|
274 |
+
uploaded_html_info = gr.HTML(label="Info")
|
275 |
+
|
276 |
+
upload_button.click(
|
277 |
+
process_uploaded_file,
|
278 |
+
inputs=[
|
279 |
+
embedding_framework_radio,
|
280 |
+
embedding_model_dropdown,
|
281 |
+
speaker_segmentation_model_dropdown,
|
282 |
+
input_num_speakers,
|
283 |
+
input_threshold,
|
284 |
+
uploaded_file,
|
285 |
+
],
|
286 |
+
outputs=[uploaded_output, uploaded_html_info],
|
287 |
+
)
|
288 |
+
|
289 |
+
gr.Markdown(description)
|
290 |
+
|
291 |
+
if __name__ == "__main__":
|
292 |
+
formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
|
293 |
+
|
294 |
+
logging.basicConfig(format=formatter, level=logging.INFO)
|
295 |
+
|
296 |
+
demo.launch()
|
model.py
ADDED
@@ -0,0 +1,152 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2024 Xiaomi Corp. (authors: Fangjun Kuang)
|
2 |
+
#
|
3 |
+
# See LICENSE for clarification regarding multiple authors
|
4 |
+
#
|
5 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
6 |
+
# you may not use this file except in compliance with the License.
|
7 |
+
# You may obtain a copy of the License at
|
8 |
+
#
|
9 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
10 |
+
#
|
11 |
+
# Unless required by applicable law or agreed to in writing, software
|
12 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
13 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
14 |
+
# See the License for the specific language governing permissions and
|
15 |
+
# limitations under the License.
|
16 |
+
|
17 |
+
import wave
|
18 |
+
from typing import List, Tuple
|
19 |
+
|
20 |
+
import numpy as np
|
21 |
+
import sherpa_onnx
|
22 |
+
from huggingface_hub import hf_hub_download
|
23 |
+
|
24 |
+
|
25 |
+
def read_wave(wave_filename: str) -> Tuple[np.ndarray, int]:
|
26 |
+
"""
|
27 |
+
Args:
|
28 |
+
wave_filename:
|
29 |
+
Path to a wave file. It should be single channel and each sample should
|
30 |
+
be 16-bit. Its sample rate does not need to be 16kHz.
|
31 |
+
Returns:
|
32 |
+
Return a tuple containing:
|
33 |
+
- A 1-D array of dtype np.float32 containing the samples, which are
|
34 |
+
normalized to the range [-1, 1].
|
35 |
+
- sample rate of the wave file
|
36 |
+
"""
|
37 |
+
|
38 |
+
with wave.open(wave_filename) as f:
|
39 |
+
assert f.getnchannels() == 1, f.getnchannels()
|
40 |
+
assert f.getsampwidth() == 2, f.getsampwidth() # it is in bytes
|
41 |
+
num_samples = f.getnframes()
|
42 |
+
samples = f.readframes(num_samples)
|
43 |
+
samples_int16 = np.frombuffer(samples, dtype=np.int16)
|
44 |
+
samples_float32 = samples_int16.astype(np.float32)
|
45 |
+
|
46 |
+
samples_float32 = samples_float32 / 32768
|
47 |
+
return samples_float32, f.getframerate()
|
48 |
+
|
49 |
+
|
50 |
+
def _get_nn_model_filename(
|
51 |
+
repo_id: str,
|
52 |
+
filename: str,
|
53 |
+
subfolder: str = ".",
|
54 |
+
) -> str:
|
55 |
+
nn_model_filename = hf_hub_download(
|
56 |
+
repo_id=repo_id,
|
57 |
+
filename=filename,
|
58 |
+
subfolder=subfolder,
|
59 |
+
)
|
60 |
+
return nn_model_filename
|
61 |
+
|
62 |
+
|
63 |
+
def get_speaker_segmentation_model(repo_id) -> List[str]:
|
64 |
+
assert repo_id in ("pyannote/segmentation-3.0",)
|
65 |
+
|
66 |
+
if repo_id == "pyannote/segmentation-3.0":
|
67 |
+
return _get_nn_model_filename(
|
68 |
+
repo_id="csukuangfj/sherpa-onnx-pyannote-segmentation-3-0",
|
69 |
+
filename="model.onnx",
|
70 |
+
)
|
71 |
+
|
72 |
+
|
73 |
+
def get_speaker_embedding_model(model_name) -> List[str]:
|
74 |
+
assert (
|
75 |
+
model_name
|
76 |
+
in three_d_speaker_embedding_models
|
77 |
+
+ nemo_speaker_embedding_models
|
78 |
+
+ wespeaker_embedding_models
|
79 |
+
)
|
80 |
+
|
81 |
+
return _get_nn_model_filename(
|
82 |
+
repo_id="csukuangfj/speaker-embedding-models",
|
83 |
+
filename=model_name,
|
84 |
+
)
|
85 |
+
|
86 |
+
|
87 |
+
def get_speaker_diarization(
|
88 |
+
segmentation_model: str, embedding_model: str, num_clusters: int, threshold: float
|
89 |
+
):
|
90 |
+
segmentation = get_speaker_segmentation_model(segmentation_model)
|
91 |
+
embedding = get_speaker_embedding_model(embedding_model)
|
92 |
+
|
93 |
+
config = sherpa_onnx.OfflineSpeakerDiarizationConfig(
|
94 |
+
segmentation=sherpa_onnx.OfflineSpeakerSegmentationModelConfig(
|
95 |
+
pyannote=sherpa_onnx.OfflineSpeakerSegmentationPyannoteModelConfig(
|
96 |
+
model=segmentation
|
97 |
+
),
|
98 |
+
),
|
99 |
+
embedding=sherpa_onnx.SpeakerEmbeddingExtractorConfig(model=embedding),
|
100 |
+
clustering=sherpa_onnx.FastClusteringConfig(
|
101 |
+
num_clusters=num_clusters,
|
102 |
+
threshold=threshold,
|
103 |
+
),
|
104 |
+
min_duration_on=0.3,
|
105 |
+
min_duration_off=0.5,
|
106 |
+
)
|
107 |
+
if not config.validate():
|
108 |
+
raise RuntimeError(
|
109 |
+
"Please check your config and make sure all required files exist"
|
110 |
+
)
|
111 |
+
|
112 |
+
return sherpa_onnx.OfflineSpeakerDiarization(config)
|
113 |
+
pass
|
114 |
+
|
115 |
+
|
116 |
+
speaker_segmentation_models = ["pyannote/segmentation-3.0"]
|
117 |
+
|
118 |
+
|
119 |
+
nemo_speaker_embedding_models = [
|
120 |
+
"nemo_en_speakerverification_speakernet.onnx",
|
121 |
+
"nemo_en_titanet_large.onnx",
|
122 |
+
"nemo_en_titanet_small.onnx",
|
123 |
+
]
|
124 |
+
|
125 |
+
three_d_speaker_embedding_models = [
|
126 |
+
"3dspeaker_speech_campplus_sv_en_voxceleb_16k.onnx",
|
127 |
+
"3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx",
|
128 |
+
"3dspeaker_speech_campplus_sv_zh_en_16k-common_advanced.onnx",
|
129 |
+
"3dspeaker_speech_eres2net_base_200k_sv_zh-cn_16k-common.onnx",
|
130 |
+
"3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx",
|
131 |
+
"3dspeaker_speech_eres2net_large_sv_zh-cn_3dspeaker_16k.onnx",
|
132 |
+
"3dspeaker_speech_eres2net_sv_en_voxceleb_16k.onnx",
|
133 |
+
"3dspeaker_speech_eres2net_sv_zh-cn_16k-common.onnx",
|
134 |
+
"3dspeaker_speech_eres2netv2_sv_zh-cn_16k-common.onnx",
|
135 |
+
]
|
136 |
+
wespeaker_embedding_models = [
|
137 |
+
"wespeaker_en_voxceleb_CAM++.onnx",
|
138 |
+
"wespeaker_en_voxceleb_CAM++_LM.onnx",
|
139 |
+
"wespeaker_en_voxceleb_resnet152_LM.onnx",
|
140 |
+
"wespeaker_en_voxceleb_resnet221_LM.onnx",
|
141 |
+
"wespeaker_en_voxceleb_resnet293_LM.onnx",
|
142 |
+
"wespeaker_en_voxceleb_resnet34.onnx",
|
143 |
+
"wespeaker_en_voxceleb_resnet34_LM.onnx",
|
144 |
+
"wespeaker_zh_cnceleb_resnet34.onnx",
|
145 |
+
"wespeaker_zh_cnceleb_resnet34_LM.onnx",
|
146 |
+
]
|
147 |
+
|
148 |
+
embedding2models = {
|
149 |
+
"3D-Speaker": three_d_speaker_embedding_models,
|
150 |
+
"NeMo": nemo_speaker_embedding_models,
|
151 |
+
"WeSpeaker": wespeaker_embedding_models,
|
152 |
+
}
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
huggingface_hub
|
2 |
+
|
3 |
+
#https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-1.9.26-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
|
4 |
+
|
5 |
+
sherpa-onnx>=1.10.28
|