Spaces:
Sleeping
Sleeping
root
commited on
Commit
•
f5c5225
1
Parent(s):
7019c9d
adding app and reqs
Browse files- app.py +32 -0
- requirements.txt +318 -0
app.py
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from pathlib import Path
|
3 |
+
import nemo
|
4 |
+
import nemo.collections.asr as nemo_asr
|
5 |
+
import librosa
|
6 |
+
import soundfile as sf
|
7 |
+
|
8 |
+
|
9 |
+
base_path = str(Path(__file__).parent)
|
10 |
+
|
11 |
+
# Converting the original wav to the same sr
|
12 |
+
def convert_wav_to_16k(input_wav_path, output_file_path, sr=16000):
|
13 |
+
y, s = librosa.load(input_wav_path, sr=sr)
|
14 |
+
sf.write(output_file_path, y, s)
|
15 |
+
print(f'"{input_wav_path}" has been converted to {s}Hz')
|
16 |
+
return output_file_path
|
17 |
+
|
18 |
+
def loading_nemo_and_prediction(processed_wav):
|
19 |
+
arabic_asr = nemo_asr.models.EncDecCTCModelBPE.restore_from(restore_path="conformer_ctc_small_60e_adamw_30wtr_32wv_40wte.nemo") # loading the model from a path
|
20 |
+
prediction = arabic_asr.transcribe(paths2audio_files=[processed_wav])
|
21 |
+
return prediction
|
22 |
+
|
23 |
+
def predict(uploaded_wav):
|
24 |
+
out_path = base_path + "/converted.wav"
|
25 |
+
audio_conversion = convert_wav_to_16k(uploaded_wav, out_path)
|
26 |
+
prediction_text = loading_nemo_and_prediction(audio_conversion)
|
27 |
+
return prediction_text[0]
|
28 |
+
|
29 |
+
|
30 |
+
|
31 |
+
demo = gr.Interface(fn=predict,inputs=gr.Audio(value='str',label="Audio file", max_length=10, show_download_button=False, interactive=True, type="filepath"), outputs=gr.Text())
|
32 |
+
demo.launch(debug=True, share=True)
|
requirements.txt
ADDED
@@ -0,0 +1,318 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
absl-py==2.1.0
|
2 |
+
addict==2.4.0
|
3 |
+
aiofiles==23.2.1
|
4 |
+
aiohttp==3.9.5
|
5 |
+
aiosignal==1.3.1
|
6 |
+
alabaster==0.7.16
|
7 |
+
altair==5.3.0
|
8 |
+
aniso8601==9.0.1
|
9 |
+
annotated-types==0.6.0
|
10 |
+
antlr4-python3-runtime==4.9.3
|
11 |
+
anyio==4.3.0
|
12 |
+
appdirs==1.4.4
|
13 |
+
asciitree==0.3.3
|
14 |
+
asttokens==2.4.1
|
15 |
+
async-timeout==4.0.3
|
16 |
+
attrdict==2.0.1
|
17 |
+
attrs==23.2.0
|
18 |
+
audioread==3.0.1
|
19 |
+
Babel==2.14.0
|
20 |
+
backcall==0.2.0
|
21 |
+
beautifulsoup4==4.12.3
|
22 |
+
black==19.10b0
|
23 |
+
bleach==6.1.0
|
24 |
+
boto3==1.34.92
|
25 |
+
botocore==1.34.92
|
26 |
+
braceexpand==0.1.7
|
27 |
+
Brotli @ file:///home/conda/feedstock_root/build_artifacts/brotli-split_1648883617327/work
|
28 |
+
cdifflib==1.2.6
|
29 |
+
certifi @ file:///home/conda/feedstock_root/build_artifacts/certifi_1707022139797/work/certifi
|
30 |
+
cffi @ file:///croot/cffi_1700254295673/work
|
31 |
+
charset-normalizer @ file:///home/conda/feedstock_root/build_artifacts/charset-normalizer_1698833585322/work
|
32 |
+
click==8.0.2
|
33 |
+
clip==0.2.0
|
34 |
+
colorama @ file:///home/conda/feedstock_root/build_artifacts/colorama_1666700638685/work
|
35 |
+
comm==0.2.2
|
36 |
+
contourpy==1.2.1
|
37 |
+
cycler==0.12.1
|
38 |
+
Cython==3.0.10
|
39 |
+
cytoolz==0.12.3
|
40 |
+
datasets==2.19.0
|
41 |
+
debugpy==1.8.1
|
42 |
+
decorator==5.1.1
|
43 |
+
defusedxml==0.7.1
|
44 |
+
diffusers==0.27.2
|
45 |
+
dill==0.3.8
|
46 |
+
Distance==0.1.3
|
47 |
+
docker-pycreds==0.4.0
|
48 |
+
docopt==0.6.2
|
49 |
+
docutils==0.21.2
|
50 |
+
editdistance==0.8.1
|
51 |
+
einops==0.7.0
|
52 |
+
einops-exts==0.0.4
|
53 |
+
exceptiongroup==1.2.1
|
54 |
+
executing==2.0.1
|
55 |
+
faiss-cpu==1.8.0
|
56 |
+
fastapi==0.110.3
|
57 |
+
fasteners==0.19
|
58 |
+
fastjsonschema==2.19.1
|
59 |
+
fasttext==0.9.2
|
60 |
+
ffmpy==0.3.2
|
61 |
+
filelock==3.13.4
|
62 |
+
Flask==2.2.5
|
63 |
+
Flask-RESTful==0.3.10
|
64 |
+
fonttools==4.51.0
|
65 |
+
frozenlist==1.4.1
|
66 |
+
fsspec @ file:///home/conda/feedstock_root/build_artifacts/fsspec_1710808267764/work
|
67 |
+
ftfy==6.2.0
|
68 |
+
future==1.0.0
|
69 |
+
g2p-en==2.1.0
|
70 |
+
gdown==5.1.0
|
71 |
+
gitdb==4.0.11
|
72 |
+
GitPython==3.1.43
|
73 |
+
google==3.0.0
|
74 |
+
gradio==4.28.3
|
75 |
+
gradio_client==0.16.0
|
76 |
+
grpcio==1.62.2
|
77 |
+
h11==0.14.0
|
78 |
+
h5py==3.11.0
|
79 |
+
httpcore==1.0.5
|
80 |
+
httpx==0.27.0
|
81 |
+
huggingface-hub==0.22.2
|
82 |
+
hydra-core==1.3.2
|
83 |
+
idna @ file:///home/conda/feedstock_root/build_artifacts/idna_1713279365350/work
|
84 |
+
ijson==3.2.3
|
85 |
+
imageio==2.34.1
|
86 |
+
imagesize==1.4.1
|
87 |
+
importlib_metadata==7.1.0
|
88 |
+
importlib_resources==6.4.0
|
89 |
+
inflect==7.2.1
|
90 |
+
iniconfig==2.0.0
|
91 |
+
intervaltree==3.1.0
|
92 |
+
ipykernel==6.29.4
|
93 |
+
ipython==8.12.3
|
94 |
+
ipywidgets==8.1.2
|
95 |
+
isort==5.13.2
|
96 |
+
itsdangerous==2.2.0
|
97 |
+
jedi==0.19.1
|
98 |
+
jieba==0.42.1
|
99 |
+
Jinja2 @ file:///home/conda/feedstock_root/build_artifacts/jinja2_1704966972576/work
|
100 |
+
jiwer==2.5.2
|
101 |
+
jmespath==1.0.1
|
102 |
+
joblib==1.4.0
|
103 |
+
jsonschema==4.21.1
|
104 |
+
jsonschema-specifications==2023.12.1
|
105 |
+
jupyter_client==8.6.1
|
106 |
+
jupyter_core==5.7.2
|
107 |
+
jupyterlab_pygments==0.3.0
|
108 |
+
jupyterlab_widgets==3.0.10
|
109 |
+
kaldi-python-io==1.2.2
|
110 |
+
kaldiio==2.18.0
|
111 |
+
kenlm @ https://github.com/kpu/kenlm/archive/master.zip#sha256=9aca61fb9df045ad86203e04b750e787403dfe4d7b86b3e99173a29f5d12d3c6
|
112 |
+
kiwisolver==1.4.5
|
113 |
+
kornia==0.7.2
|
114 |
+
kornia_rs==0.1.3
|
115 |
+
latexcodec==3.0.0
|
116 |
+
lazy_loader==0.4
|
117 |
+
Levenshtein==0.22.0
|
118 |
+
lhotse==1.22.0
|
119 |
+
librosa==0.10.1
|
120 |
+
lightning @ file:///home/conda/feedstock_root/build_artifacts/lightning_1713989954271/work
|
121 |
+
lightning-utilities @ file:///home/conda/feedstock_root/build_artifacts/lightning-utilities_1711597355069/work
|
122 |
+
lilcom==1.7
|
123 |
+
llvmlite==0.42.0
|
124 |
+
loguru==0.7.2
|
125 |
+
lxml==5.2.1
|
126 |
+
Markdown==3.6
|
127 |
+
markdown-it-py==3.0.0
|
128 |
+
markdown2==2.4.13
|
129 |
+
MarkupSafe @ file:///croot/markupsafe_1704205993651/work
|
130 |
+
marshmallow==3.21.1
|
131 |
+
matplotlib==3.8.4
|
132 |
+
matplotlib-inline==0.1.7
|
133 |
+
mdurl==0.1.2
|
134 |
+
megatron_core==0.5.0
|
135 |
+
mistune==3.0.2
|
136 |
+
mkl-fft @ file:///croot/mkl_fft_1695058164594/work
|
137 |
+
mkl-random @ file:///croot/mkl_random_1695059800811/work
|
138 |
+
mkl-service==2.4.0
|
139 |
+
more-itertools==10.2.0
|
140 |
+
mpmath @ file:///home/conda/feedstock_root/build_artifacts/mpmath_1678228039184/work
|
141 |
+
msgpack==1.0.8
|
142 |
+
multidict==6.0.5
|
143 |
+
multiprocess==0.70.16
|
144 |
+
nbclient==0.10.0
|
145 |
+
nbconvert==7.16.4
|
146 |
+
nbformat==5.10.4
|
147 |
+
nemo_text_processing==0.3.0rc0
|
148 |
+
nemo_toolkit==1.23.0
|
149 |
+
nerfacc==0.5.3
|
150 |
+
nest-asyncio==1.6.0
|
151 |
+
networkx @ file:///home/conda/feedstock_root/build_artifacts/networkx_1712540363324/work
|
152 |
+
nltk==3.8.1
|
153 |
+
numba==0.59.1
|
154 |
+
numcodecs==0.12.1
|
155 |
+
numpy @ file:///croot/numpy_and_numpy_base_1708638617955/work/dist/numpy-1.26.4-cp310-cp310-linux_x86_64.whl#sha256=d8cd837ed43e87f77e6efaa08e8de927ca030a1c9c5d04624432d6fb9a74a5ee
|
156 |
+
nvidia-cublas-cu12==12.1.3.1
|
157 |
+
nvidia-cuda-cupti-cu12==12.1.105
|
158 |
+
nvidia-cuda-nvrtc-cu12==12.1.105
|
159 |
+
nvidia-cuda-runtime-cu12==12.1.105
|
160 |
+
nvidia-cudnn-cu12==8.9.2.26
|
161 |
+
nvidia-cufft-cu12==11.0.2.54
|
162 |
+
nvidia-curand-cu12==10.3.2.106
|
163 |
+
nvidia-cusolver-cu12==11.4.5.107
|
164 |
+
nvidia-cusparse-cu12==12.1.0.106
|
165 |
+
nvidia-nccl-cu12==2.20.5
|
166 |
+
nvidia-nvjitlink-cu12==12.4.127
|
167 |
+
nvidia-nvtx-cu12==12.1.105
|
168 |
+
omegaconf==2.3.0
|
169 |
+
onnx==1.16.0
|
170 |
+
open-clip-torch==2.24.0
|
171 |
+
OpenCC==1.1.6
|
172 |
+
orjson==3.10.1
|
173 |
+
packaging==24.0
|
174 |
+
pandas==2.2.2
|
175 |
+
pandocfilters==1.5.1
|
176 |
+
pangu==4.0.6.1
|
177 |
+
parameterized==0.9.0
|
178 |
+
parso==0.8.4
|
179 |
+
pathspec==0.12.1
|
180 |
+
pexpect==4.9.0
|
181 |
+
pickleshare==0.7.5
|
182 |
+
pillow==10.3.0
|
183 |
+
pipreqs==0.5.0
|
184 |
+
plac==1.4.3
|
185 |
+
platformdirs==4.2.1
|
186 |
+
pluggy==1.5.0
|
187 |
+
pooch==1.8.1
|
188 |
+
portalocker==2.8.2
|
189 |
+
progress==1.6
|
190 |
+
prompt-toolkit==3.0.43
|
191 |
+
protobuf==4.25.3
|
192 |
+
psutil==5.9.8
|
193 |
+
ptyprocess==0.7.0
|
194 |
+
pure-eval==0.2.2
|
195 |
+
pyannote.core==5.0.0
|
196 |
+
pyannote.database==5.1.0
|
197 |
+
pyannote.metrics==3.2.1
|
198 |
+
PyArabic==0.6.15
|
199 |
+
pyarrow==16.0.0
|
200 |
+
pyarrow-hotfix==0.6
|
201 |
+
PyAudio==0.2.14
|
202 |
+
pybind11==2.12.0
|
203 |
+
pybtex==0.24.0
|
204 |
+
pybtex-docutils==1.0.3
|
205 |
+
pycparser @ file:///home/conda/feedstock_root/build_artifacts/pycparser_1711811537435/work
|
206 |
+
pydantic==2.7.1
|
207 |
+
pydantic_core==2.18.2
|
208 |
+
pydub==0.25.1
|
209 |
+
Pygments==2.17.2
|
210 |
+
pyloudnorm==0.1.1
|
211 |
+
PyMCubes==0.1.4
|
212 |
+
pynini==2.1.5
|
213 |
+
pyparsing==3.1.2
|
214 |
+
pypinyin==0.51.0
|
215 |
+
pypinyin-dict==0.8.0
|
216 |
+
PySocks @ file:///home/conda/feedstock_root/build_artifacts/pysocks_1661604839144/work
|
217 |
+
pytest==8.1.1
|
218 |
+
pytest-runner==6.0.1
|
219 |
+
python-dateutil==2.9.0.post0
|
220 |
+
python-multipart==0.0.9
|
221 |
+
pytorch-lightning==2.0.7
|
222 |
+
pytz==2024.1
|
223 |
+
PyYAML==6.0.1
|
224 |
+
pyzmq==26.0.2
|
225 |
+
rapidfuzz==2.13.7
|
226 |
+
referencing==0.35.0
|
227 |
+
regex==2024.4.16
|
228 |
+
requests @ file:///home/conda/feedstock_root/build_artifacts/requests_1684774241324/work
|
229 |
+
resampy==0.4.3
|
230 |
+
rich==13.7.1
|
231 |
+
rouge-score==0.1.2
|
232 |
+
rpds-py==0.18.0
|
233 |
+
ruamel.yaml==0.18.6
|
234 |
+
ruamel.yaml.clib==0.2.8
|
235 |
+
ruff==0.4.2
|
236 |
+
s3transfer==0.10.1
|
237 |
+
sacrebleu==2.4.2
|
238 |
+
sacremoses==0.1.1
|
239 |
+
safetensors==0.4.3
|
240 |
+
scikit-learn==1.4.2
|
241 |
+
scipy==1.13.0
|
242 |
+
semantic-version==2.10.0
|
243 |
+
sentence-transformers==2.7.0
|
244 |
+
sentencepiece==0.2.0
|
245 |
+
sentry-sdk==2.0.0
|
246 |
+
setproctitle==1.3.3
|
247 |
+
shellingham==1.5.4
|
248 |
+
six==1.16.0
|
249 |
+
smmap==5.0.1
|
250 |
+
sniffio==1.3.1
|
251 |
+
snowballstemmer==2.2.0
|
252 |
+
sortedcontainers==2.4.0
|
253 |
+
soundfile==0.12.1
|
254 |
+
soupsieve==2.5
|
255 |
+
sox==1.5.0
|
256 |
+
soxr==0.3.7
|
257 |
+
Sphinx==7.3.7
|
258 |
+
sphinxcontrib-applehelp==1.0.8
|
259 |
+
sphinxcontrib-bibtex==2.6.2
|
260 |
+
sphinxcontrib-devhelp==1.0.6
|
261 |
+
sphinxcontrib-htmlhelp==2.0.5
|
262 |
+
sphinxcontrib-jsmath==1.0.1
|
263 |
+
sphinxcontrib-qthelp==1.0.7
|
264 |
+
sphinxcontrib-serializinghtml==1.1.10
|
265 |
+
stack-data==0.6.3
|
266 |
+
starlette==0.37.2
|
267 |
+
sympy @ file:///home/conda/feedstock_root/build_artifacts/sympy_1684180539862/work
|
268 |
+
tabulate==0.9.0
|
269 |
+
taming-transformers==0.0.1
|
270 |
+
tensorboard==2.16.2
|
271 |
+
tensorboard-data-server==0.7.2
|
272 |
+
tensorstore==0.1.45
|
273 |
+
termcolor==2.4.0
|
274 |
+
text-unidecode==1.3
|
275 |
+
textdistance==4.6.2
|
276 |
+
texterrors==0.4.4
|
277 |
+
threadpoolctl==3.4.0
|
278 |
+
timm==0.9.16
|
279 |
+
tinycss2==1.3.0
|
280 |
+
tokenizers==0.19.1
|
281 |
+
toml==0.10.2
|
282 |
+
tomli==2.0.1
|
283 |
+
tomlkit==0.12.0
|
284 |
+
toolz==0.12.1
|
285 |
+
torch==2.3.0
|
286 |
+
torchdiffeq==0.2.3
|
287 |
+
torchmetrics @ file:///home/conda/feedstock_root/build_artifacts/torchmetrics_1710773712978/work
|
288 |
+
torchsde==0.2.6
|
289 |
+
torchvision==0.18.0
|
290 |
+
tornado==6.4
|
291 |
+
tqdm @ file:///home/conda/feedstock_root/build_artifacts/tqdm_1707598593068/work
|
292 |
+
traitlets==5.14.3
|
293 |
+
trampoline==0.1.2
|
294 |
+
transformers==4.40.1
|
295 |
+
trimesh==4.3.1
|
296 |
+
triton==2.3.0
|
297 |
+
typed-ast==1.5.5
|
298 |
+
typeguard==4.2.1
|
299 |
+
typer==0.12.3
|
300 |
+
typing_extensions==4.11.0
|
301 |
+
tzdata==2024.1
|
302 |
+
urllib3 @ file:///home/conda/feedstock_root/build_artifacts/urllib3_1708239446578/work
|
303 |
+
uvicorn==0.29.0
|
304 |
+
wandb==0.16.6
|
305 |
+
wcwidth==0.2.13
|
306 |
+
webdataset==0.1.62
|
307 |
+
webencodings==0.5.1
|
308 |
+
websockets==11.0.3
|
309 |
+
Werkzeug==3.0.2
|
310 |
+
wget==3.2
|
311 |
+
widgetsnbextension==4.0.10
|
312 |
+
wrapt==1.16.0
|
313 |
+
xxhash==3.4.1
|
314 |
+
yarg==0.1.9
|
315 |
+
yarl==1.9.4
|
316 |
+
youtokentome==1.0.6
|
317 |
+
zarr==2.17.2
|
318 |
+
zipp==3.18.1
|