Spaces:
Sleeping
Sleeping
File size: 9,313 Bytes
5ce1fe8 7e1376c 168b5c0 bb7ea32 168b5c0 dcf6c65 168b5c0 5ce1fe8 168b5c0 5ce1fe8 168b5c0 3e60665 168b5c0 5ce1fe8 168b5c0 5ce1fe8 168b5c0 5ce1fe8 3e60665 2267fac 3e60665 2267fac 5ce1fe8 2267fac 168b5c0 dcf6c65 1b1a336 dcf6c65 3e60665 168b5c0 dcf6c65 168b5c0 085d464 3e60665 168b5c0 085d464 168b5c0 dcf6c65 168b5c0 a1e736d 168b5c0 dcf6c65 a1e736d 168b5c0 2267fac 168b5c0 2267fac 168b5c0 5ce1fe8 e7f863b 5ce1fe8 168b5c0 5ce1fe8 3e60665 5ce1fe8 7e1376c 3e60665 7e1376c 5ce1fe8 e7f863b 5ce1fe8 e7f863b 5ce1fe8 168b5c0 5ce1fe8 f1ed46a 168b5c0 f1ed46a efb9bdb bb7ea32 5ce1fe8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 |
#!/usr/bin/python3
# -*- coding: utf-8 -*-
import argparse
from collections import defaultdict
from datetime import datetime
import functools
import io
import logging
from pathlib import Path
import platform
import time
import tempfile
from project_settings import project_path, log_directory
import log
log.setup(log_directory=log_directory)
import gradio as gr
import torch
import torchaudio
from toolbox.k2_sherpa.examples import examples
from toolbox.k2_sherpa import decode, nn_models
from toolbox.k2_sherpa.utils import audio_convert
main_logger = logging.getLogger("main")
def get_args():
parser = argparse.ArgumentParser()
parser.add_argument(
"--pretrained_model_dir",
default=(project_path / "pretrained_models").as_posix(),
type=str
)
args = parser.parse_args()
return args
def update_model_dropdown(language: str):
if language not in nn_models.model_map.keys():
raise ValueError(f"Unsupported language: {language}")
choices = nn_models.model_map[language]
choices = [c["repo_id"] for c in choices]
return gr.Dropdown(
choices=choices,
value=choices[0],
interactive=True,
)
def build_html_output(s: str, style: str = "result_item_success"):
return f"""
<div class='result'>
<div class='result_item {style}'>
{s}
</div>
</div>
"""
@torch.no_grad()
def process(
language: str,
repo_id: str,
decoding_method: str,
num_active_paths: int,
add_punctuation: str,
in_filename: str,
pretrained_model_dir: Path,
):
main_logger.info("language: {}".format(language))
main_logger.info("repo_id: {}".format(repo_id))
main_logger.info("decoding_method: {}".format(decoding_method))
main_logger.info("num_active_paths: {}".format(num_active_paths))
main_logger.info("in_filename: {}".format(in_filename))
# audio convert
in_filename = Path(in_filename)
out_filename = Path(tempfile.gettempdir()) / "asr" / in_filename.name
out_filename.parent.mkdir(parents=True, exist_ok=True)
audio_convert(in_filename=in_filename.as_posix(),
out_filename=out_filename.as_posix(),
)
# model settings
m_list = nn_models.model_map.get(language)
if m_list is None:
raise AssertionError("language invalid: {}".format(language))
m_dict = None
for m in m_list:
if m["repo_id"] == repo_id:
m_dict = m
if m_dict is None:
raise AssertionError("repo_id invalid: {}".format(repo_id))
# load recognizer
local_model_dir = pretrained_model_dir / "huggingface" / repo_id
nn_model_file = local_model_dir / m_dict["nn_model_file"]
tokens_file = local_model_dir / m_dict["tokens_file"]
recognizer = nn_models.load_recognizer(
repo_id=m_dict["repo_id"],
nn_model_file=nn_model_file.as_posix(),
tokens_file=tokens_file.as_posix(),
sub_folder=m_dict["sub_folder"],
local_model_dir=local_model_dir,
recognizer_type=m_dict["recognizer_type"],
decoding_method=decoding_method,
num_active_paths=num_active_paths,
)
# transcribe
now = datetime.now()
date_time = now.strftime("%Y-%m-%d %H:%M:%S.%f")
logging.info(f"Started at {date_time}")
start = time.time()
text = decode.decode_by_recognizer(recognizer=recognizer,
filename=out_filename.as_posix(),
)
date_time = now.strftime("%Y-%m-%d %H:%M:%S.%f")
end = time.time()
# statistics
metadata = torchaudio.info(out_filename.as_posix())
duration = metadata.num_frames / 16000
rtf = (end - start) / duration
main_logger.info(f"Finished at {date_time} s. Elapsed: {end - start: .3f} s")
info = f"""
Wave duration : {duration: .3f} s <br/>
Processing time: {end - start: .3f} s <br/>
RTF: {end - start: .3f}/{duration: .3f} = {rtf:.3f} <br/>
"""
main_logger.info(info)
main_logger.info(f"\nrepo_id: {repo_id}\nhyp: {text}")
return text, build_html_output(info)
def process_uploaded_file(language: str,
repo_id: str,
decoding_method: str,
num_active_paths: int,
add_punctuation: str,
in_filename: str,
pretrained_model_dir: Path,
):
if in_filename is None or in_filename == "":
return "", build_html_output(
"Please first upload a file and then click "
'the button "submit for recognition"',
"result_item_error",
)
main_logger.info(f"Processing uploaded file: {in_filename}")
try:
return process(
in_filename=in_filename,
language=language,
repo_id=repo_id,
decoding_method=decoding_method,
num_active_paths=num_active_paths,
add_punctuation=add_punctuation,
pretrained_model_dir=pretrained_model_dir,
)
except Exception as e:
msg = "transcribe error: {}".format(str(e))
main_logger.info(msg)
return "", build_html_output(msg, "result_item_error")
# css style is copied from
# https://huggingface.co/spaces/alphacep/asr/blob/main/app.py#L113
css = """
.result {display:flex;flex-direction:column}
.result_item {padding:15px;margin-bottom:8px;border-radius:15px;width:100%}
.result_item_success {background-color:mediumaquamarine;color:white;align-self:start}
.result_item_error {background-color:#ff7070;color:white;align-self:start}
"""
def main():
args = get_args()
pretrained_model_dir = Path(args.pretrained_model_dir)
pretrained_model_dir.mkdir(exist_ok=True)
process_uploaded_file_ = functools.partial(
process_uploaded_file,
pretrained_model_dir=pretrained_model_dir,
)
title = "# Automatic Speech Recognition with Next-gen Kaldi"
language_choices = list(nn_models.model_map.keys())
language_to_models = defaultdict(list)
for k, v in nn_models.model_map.items():
for m in v:
repo_id = m["repo_id"]
language_to_models[k].append(repo_id)
# blocks
with gr.Blocks(css=css) as blocks:
gr.Markdown(value=title)
with gr.Tabs():
with gr.TabItem("Upload from disk"):
language_radio = gr.Radio(
label="Language",
choices=language_choices,
value=language_choices[0],
)
model_dropdown = gr.Dropdown(
choices=language_to_models[language_choices[0]],
label="Select a model",
value=language_to_models[language_choices[0]][0],
)
decoding_method_radio = gr.Radio(
label="Decoding method",
choices=["greedy_search", "modified_beam_search"],
value="greedy_search",
)
num_active_paths_slider = gr.Slider(
minimum=1,
value=4,
step=1,
label="Number of active paths for modified_beam_search",
)
punct_radio = gr.Radio(
label="Whether to add punctuation (Only for Chinese and English)",
choices=["Yes", "No"],
value="Yes",
)
uploaded_file = gr.Audio(
sources=["upload"],
type="filepath",
label="Upload from disk",
)
upload_button = gr.Button("Submit for recognition")
uploaded_output = gr.Textbox(label="Recognized speech from uploaded file")
uploaded_html_info = gr.HTML(label="Info")
gr.Examples(
examples=examples,
inputs=[
language_radio,
model_dropdown,
decoding_method_radio,
num_active_paths_slider,
punct_radio,
uploaded_file,
],
outputs=[uploaded_output, uploaded_html_info],
fn=process_uploaded_file_,
)
upload_button.click(
process_uploaded_file_,
inputs=[
language_radio,
model_dropdown,
decoding_method_radio,
num_active_paths_slider,
punct_radio,
uploaded_file,
],
outputs=[uploaded_output, uploaded_html_info],
)
language_radio.change(
update_model_dropdown,
inputs=language_radio,
outputs=model_dropdown,
)
blocks.queue().launch(
share=False if platform.system() == "Windows" else False,
server_name="127.0.0.1" if platform.system() == "Windows" else "0.0.0.0",
server_port=7860
)
return
if __name__ == "__main__":
main()
|