vunhucuongit
commited on
Update soni_translate/speech_segmentation.py
Browse files
soni_translate/speech_segmentation.py
CHANGED
@@ -9,7 +9,6 @@ import gc
|
|
9 |
import os
|
10 |
import soundfile as sf
|
11 |
from IPython.utils import capture # noqa
|
12 |
-
from .language_configuration import EXTRA_ALIGN, INVERTED_LANGUAGES
|
13 |
from .logging_setup import logger
|
14 |
from .postprocessor import sanitize_file_name
|
15 |
from .utils import remove_directory_contents, run_command
|
@@ -59,8 +58,6 @@ def load_align_and_align_segments(result, audio, DAMHF):
|
|
59 |
language_code=result["language"],
|
60 |
device=os.environ.get("SONITR_DEVICE") if os.environ.get("ZERO_GPU") != "TRUE" else "cuda",
|
61 |
model_name=None
|
62 |
-
if result["language"] in DAMHF.keys()
|
63 |
-
else EXTRA_ALIGN[result["language"]],
|
64 |
)
|
65 |
|
66 |
# Align segments
|
@@ -343,31 +340,6 @@ def align_speech(audio, result):
|
|
343 |
- Cleans up memory by releasing resources after alignment.
|
344 |
"""
|
345 |
DAMHF.update(DAMT) # lang align
|
346 |
-
if (
|
347 |
-
not result["language"] in DAMHF.keys()
|
348 |
-
and not result["language"] in EXTRA_ALIGN.keys()
|
349 |
-
):
|
350 |
-
logger.warning(
|
351 |
-
"Automatic detection: Source language not compatible with align"
|
352 |
-
)
|
353 |
-
raise ValueError(
|
354 |
-
f"Detected language {result['language']} incompatible, "
|
355 |
-
"you can select the source language to avoid this error."
|
356 |
-
)
|
357 |
-
if (
|
358 |
-
result["language"] in EXTRA_ALIGN.keys()
|
359 |
-
and EXTRA_ALIGN[result["language"]] == ""
|
360 |
-
):
|
361 |
-
lang_name = (
|
362 |
-
INVERTED_LANGUAGES[result["language"]]
|
363 |
-
if result["language"] in INVERTED_LANGUAGES.keys()
|
364 |
-
else result["language"]
|
365 |
-
)
|
366 |
-
logger.warning(
|
367 |
-
"No compatible wav2vec2 model found "
|
368 |
-
f"for the language '{lang_name}', skipping alignment."
|
369 |
-
)
|
370 |
-
return result
|
371 |
|
372 |
# random_sleep()
|
373 |
result = load_align_and_align_segments(result, audio, DAMHF)
|
|
|
9 |
import os
|
10 |
import soundfile as sf
|
11 |
from IPython.utils import capture # noqa
|
|
|
12 |
from .logging_setup import logger
|
13 |
from .postprocessor import sanitize_file_name
|
14 |
from .utils import remove_directory_contents, run_command
|
|
|
58 |
language_code=result["language"],
|
59 |
device=os.environ.get("SONITR_DEVICE") if os.environ.get("ZERO_GPU") != "TRUE" else "cuda",
|
60 |
model_name=None
|
|
|
|
|
61 |
)
|
62 |
|
63 |
# Align segments
|
|
|
340 |
- Cleans up memory by releasing resources after alignment.
|
341 |
"""
|
342 |
DAMHF.update(DAMT) # lang align
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
343 |
|
344 |
# random_sleep()
|
345 |
result = load_align_and_align_segments(result, audio, DAMHF)
|