Spaces:
Running
Running
jhj0517
commited on
Commit
·
79e6f08
1
Parent(s):
a1b32c1
Refactor translation
Browse files- modules/translation/deepl_api.py +17 -26
- modules/translation/translation_base.py +18 -26
- modules/utils/subtitle_manager.py +62 -69
modules/translation/deepl_api.py
CHANGED
@@ -139,37 +139,28 @@ class DeepLAPI:
|
|
139 |
)
|
140 |
|
141 |
files_info = {}
|
142 |
-
for
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
if file_ext == ".srt":
|
147 |
-
parsed_dicts = parse_srt(file_path=file_path)
|
148 |
-
|
149 |
-
elif file_ext == ".vtt":
|
150 |
-
parsed_dicts = parse_vtt(file_path=file_path)
|
151 |
|
152 |
batch_size = self.max_text_batch_size
|
153 |
-
for batch_start in range(0, len(
|
154 |
-
|
155 |
-
sentences_to_translate = [
|
156 |
translated_texts = self.request_deepl_translate(auth_key, sentences_to_translate, source_lang,
|
157 |
target_lang, is_pro)
|
158 |
for i, translated_text in enumerate(translated_texts):
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
file_name += f"-{timestamp}"
|
170 |
-
|
171 |
-
output_path = os.path.join(self.output_dir, f"{file_name}{file_ext}")
|
172 |
-
write_file(subtitle, output_path)
|
173 |
|
174 |
files_info[file_name] = {"subtitle": subtitle, "path": output_path}
|
175 |
|
|
|
139 |
)
|
140 |
|
141 |
files_info = {}
|
142 |
+
for file_path in fileobjs:
|
143 |
+
file_name, file_ext = os.path.splitext(os.path.basename(file_path))
|
144 |
+
writer = get_writer(file_ext, self.output_dir)
|
145 |
+
segments = writer.to_segments(file_path)
|
|
|
|
|
|
|
|
|
|
|
146 |
|
147 |
batch_size = self.max_text_batch_size
|
148 |
+
for batch_start in range(0, len(segments), batch_size):
|
149 |
+
progress(batch_start / len(segments), desc="Translating..")
|
150 |
+
sentences_to_translate = segments[batch_start:batch_start+batch_size]
|
151 |
translated_texts = self.request_deepl_translate(auth_key, sentences_to_translate, source_lang,
|
152 |
target_lang, is_pro)
|
153 |
for i, translated_text in enumerate(translated_texts):
|
154 |
+
segments[batch_start + i].text = translated_text["text"]
|
155 |
+
print("DeepL Segments: ", segments)
|
156 |
+
|
157 |
+
subtitle, output_path = generate_file(
|
158 |
+
output_dir=self.output_dir,
|
159 |
+
output_file_name=file_name,
|
160 |
+
output_format=file_ext,
|
161 |
+
result=segments,
|
162 |
+
add_timestamp=add_timestamp
|
163 |
+
)
|
|
|
|
|
|
|
|
|
164 |
|
165 |
files_info[file_name] = {"subtitle": subtitle, "path": output_path}
|
166 |
|
modules/translation/translation_base.py
CHANGED
@@ -95,32 +95,22 @@ class TranslationBase(ABC):
|
|
95 |
files_info = {}
|
96 |
for fileobj in fileobjs:
|
97 |
file_name, file_ext = os.path.splitext(os.path.basename(fileobj))
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
subtitle = get_serialized_vtt(parsed_dicts)
|
115 |
-
|
116 |
-
if add_timestamp:
|
117 |
-
timestamp = datetime.now().strftime("%m%d%H%M%S")
|
118 |
-
file_name += f"-{timestamp}"
|
119 |
-
|
120 |
-
output_path = os.path.join(self.output_dir, f"{file_name}{file_ext}")
|
121 |
-
write_file(subtitle, output_path)
|
122 |
-
|
123 |
-
files_info[file_name] = {"subtitle": subtitle, "path": output_path}
|
124 |
|
125 |
total_result = ''
|
126 |
for file_name, info in files_info.items():
|
@@ -134,6 +124,8 @@ class TranslationBase(ABC):
|
|
134 |
|
135 |
except Exception as e:
|
136 |
print(f"Error: {str(e)}")
|
|
|
|
|
137 |
finally:
|
138 |
self.release_cuda_memory()
|
139 |
|
|
|
95 |
files_info = {}
|
96 |
for fileobj in fileobjs:
|
97 |
file_name, file_ext = os.path.splitext(os.path.basename(fileobj))
|
98 |
+
writer = get_writer(file_ext, self.output_dir)
|
99 |
+
segments = writer.to_segments(fileobj)
|
100 |
+
for i, segment in enumerate(segments):
|
101 |
+
progress(i / len(segments), desc="Translating..")
|
102 |
+
translated_text = self.translate(segment.text, max_length=max_length)
|
103 |
+
segment.text = translated_text
|
104 |
+
|
105 |
+
subtitle, file_path = generate_file(
|
106 |
+
output_dir=self.output_dir,
|
107 |
+
output_file_name=file_name,
|
108 |
+
output_format=file_ext,
|
109 |
+
result=segments,
|
110 |
+
add_timestamp=add_timestamp
|
111 |
+
)
|
112 |
+
|
113 |
+
files_info[file_name] = {"subtitle": subtitle, "path": file_path}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
114 |
|
115 |
total_result = ''
|
116 |
for file_name, info in files_info.items():
|
|
|
124 |
|
125 |
except Exception as e:
|
126 |
print(f"Error: {str(e)}")
|
127 |
+
import traceback
|
128 |
+
traceback.print_exc()
|
129 |
finally:
|
130 |
self.release_cuda_memory()
|
131 |
|
modules/utils/subtitle_manager.py
CHANGED
@@ -33,6 +33,18 @@ def format_timestamp(
|
|
33 |
)
|
34 |
|
35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
def get_start(segments: List[dict]) -> Optional[float]:
|
37 |
return next(
|
38 |
(w["start"] for s in segments for w in s["words"]),
|
@@ -54,16 +66,12 @@ class ResultWriter:
|
|
54 |
self.output_dir = output_dir
|
55 |
|
56 |
def __call__(
|
57 |
-
self, result: Union[dict, List[Segment]], output_file_name: str,
|
58 |
options: Optional[dict] = None, **kwargs
|
59 |
):
|
60 |
if isinstance(result, List) and result and isinstance(result[0], Segment):
|
61 |
result = {"segments": [seg.dict() for seg in result]}
|
62 |
|
63 |
-
if add_timestamp:
|
64 |
-
timestamp = datetime.now().strftime("%m%d%H%M%S")
|
65 |
-
output_file_name += f"-{timestamp}"
|
66 |
-
|
67 |
output_path = os.path.join(
|
68 |
self.output_dir, output_file_name + "." + self.extension
|
69 |
)
|
@@ -216,6 +224,26 @@ class WriteVTT(SubtitlesWriter):
|
|
216 |
for start, end, text in self.iterate_result(result, options, **kwargs):
|
217 |
print(f"{start} --> {end}\n{text}\n", file=file, flush=True)
|
218 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
219 |
|
220 |
class WriteSRT(SubtitlesWriter):
|
221 |
extension: str = "srt"
|
@@ -230,6 +258,27 @@ class WriteSRT(SubtitlesWriter):
|
|
230 |
):
|
231 |
print(f"{i}\n{start} --> {end}\n{text}\n", file=file, flush=True)
|
232 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
233 |
|
234 |
class WriteTSV(ResultWriter):
|
235 |
"""
|
@@ -265,7 +314,7 @@ class WriteJSON(ResultWriter):
|
|
265 |
def get_writer(
|
266 |
output_format: str, output_dir: str
|
267 |
) -> Callable[[dict, TextIO, dict], None]:
|
268 |
-
output_format = output_format.strip().lower()
|
269 |
|
270 |
writers = {
|
271 |
"txt": WriteTXT,
|
@@ -292,75 +341,19 @@ def get_writer(
|
|
292 |
def generate_file(
|
293 |
output_format: str, output_dir: str, result: Union[dict, List[Segment]], output_file_name: str, add_timestamp: bool = True,
|
294 |
) -> Tuple[str, str]:
|
|
|
|
|
|
|
|
|
|
|
|
|
295 |
file_path = os.path.join(output_dir, f"{output_file_name}.{output_format}")
|
296 |
file_writer = get_writer(output_format=output_format, output_dir=output_dir)
|
297 |
-
file_writer(result=result, output_file_name=output_file_name
|
298 |
content = read_file(file_path)
|
299 |
return content, file_path
|
300 |
|
301 |
|
302 |
-
def parse_srt(file_path):
|
303 |
-
"""Reads SRT file and returns as dict"""
|
304 |
-
with open(file_path, 'r', encoding='utf-8') as file:
|
305 |
-
srt_data = file.read()
|
306 |
-
|
307 |
-
data = []
|
308 |
-
blocks = srt_data.split('\n\n')
|
309 |
-
|
310 |
-
for block in blocks:
|
311 |
-
if block.strip() != '':
|
312 |
-
lines = block.strip().split('\n')
|
313 |
-
index = lines[0]
|
314 |
-
timestamp = lines[1]
|
315 |
-
sentence = ' '.join(lines[2:])
|
316 |
-
|
317 |
-
data.append({
|
318 |
-
"index": index,
|
319 |
-
"timestamp": timestamp,
|
320 |
-
"sentence": sentence
|
321 |
-
})
|
322 |
-
return data
|
323 |
-
|
324 |
-
|
325 |
-
def parse_vtt(file_path):
|
326 |
-
"""Reads WEBVTT file and returns as dict"""
|
327 |
-
with open(file_path, 'r', encoding='utf-8') as file:
|
328 |
-
webvtt_data = file.read()
|
329 |
-
|
330 |
-
data = []
|
331 |
-
blocks = webvtt_data.split('\n\n')
|
332 |
-
|
333 |
-
for block in blocks:
|
334 |
-
if block.strip() != '' and not block.strip().startswith("WEBVTT"):
|
335 |
-
lines = block.strip().split('\n')
|
336 |
-
timestamp = lines[0]
|
337 |
-
sentence = ' '.join(lines[1:])
|
338 |
-
|
339 |
-
data.append({
|
340 |
-
"timestamp": timestamp,
|
341 |
-
"sentence": sentence
|
342 |
-
})
|
343 |
-
|
344 |
-
return data
|
345 |
-
|
346 |
-
|
347 |
-
def get_serialized_srt(dicts):
|
348 |
-
output = ""
|
349 |
-
for dic in dicts:
|
350 |
-
output += f'{dic["index"]}\n'
|
351 |
-
output += f'{dic["timestamp"]}\n'
|
352 |
-
output += f'{dic["sentence"]}\n\n'
|
353 |
-
return output
|
354 |
-
|
355 |
-
|
356 |
-
def get_serialized_vtt(dicts):
|
357 |
-
output = "WEBVTT\n\n"
|
358 |
-
for dic in dicts:
|
359 |
-
output += f'{dic["timestamp"]}\n'
|
360 |
-
output += f'{dic["sentence"]}\n\n'
|
361 |
-
return output
|
362 |
-
|
363 |
-
|
364 |
def safe_filename(name):
|
365 |
INVALID_FILENAME_CHARS = r'[<>:"/\\|?*\x00-\x1f]'
|
366 |
safe_name = re.sub(INVALID_FILENAME_CHARS, '_', name)
|
|
|
33 |
)
|
34 |
|
35 |
|
36 |
+
def time_str_to_seconds(time_str: str, decimal_marker: str = ",") -> float:
|
37 |
+
hours, minutes, rest = time_str.split(":")
|
38 |
+
seconds, fractional = rest.split(decimal_marker)
|
39 |
+
|
40 |
+
hours = int(hours)
|
41 |
+
minutes = int(minutes)
|
42 |
+
seconds = int(seconds)
|
43 |
+
fractional_seconds = float("0." + fractional)
|
44 |
+
|
45 |
+
return hours * 3600 + minutes * 60 + seconds + fractional_seconds
|
46 |
+
|
47 |
+
|
48 |
def get_start(segments: List[dict]) -> Optional[float]:
|
49 |
return next(
|
50 |
(w["start"] for s in segments for w in s["words"]),
|
|
|
66 |
self.output_dir = output_dir
|
67 |
|
68 |
def __call__(
|
69 |
+
self, result: Union[dict, List[Segment]], output_file_name: str,
|
70 |
options: Optional[dict] = None, **kwargs
|
71 |
):
|
72 |
if isinstance(result, List) and result and isinstance(result[0], Segment):
|
73 |
result = {"segments": [seg.dict() for seg in result]}
|
74 |
|
|
|
|
|
|
|
|
|
75 |
output_path = os.path.join(
|
76 |
self.output_dir, output_file_name + "." + self.extension
|
77 |
)
|
|
|
224 |
for start, end, text in self.iterate_result(result, options, **kwargs):
|
225 |
print(f"{start} --> {end}\n{text}\n", file=file, flush=True)
|
226 |
|
227 |
+
def to_segments(self, file_path: str) -> List[Segment]:
|
228 |
+
segments = []
|
229 |
+
|
230 |
+
blocks = read_file(file_path).split('\n\n')
|
231 |
+
|
232 |
+
for block in blocks:
|
233 |
+
if block.strip() != '' and not block.strip().startswith("WEBVTT"):
|
234 |
+
lines = block.strip().split('\n')
|
235 |
+
time_line = lines[0].split(" --> ")
|
236 |
+
start, end = time_str_to_seconds(time_line[0], self.decimal_marker), time_str_to_seconds(time_line[1], self.decimal_marker)
|
237 |
+
sentence = ' '.join(lines[1:])
|
238 |
+
|
239 |
+
segments.append(Segment(
|
240 |
+
start=start,
|
241 |
+
end=end,
|
242 |
+
text=sentence
|
243 |
+
))
|
244 |
+
|
245 |
+
return segments
|
246 |
+
|
247 |
|
248 |
class WriteSRT(SubtitlesWriter):
|
249 |
extension: str = "srt"
|
|
|
258 |
):
|
259 |
print(f"{i}\n{start} --> {end}\n{text}\n", file=file, flush=True)
|
260 |
|
261 |
+
def to_segments(self, file_path: str) -> List[Segment]:
|
262 |
+
segments = []
|
263 |
+
|
264 |
+
blocks = read_file(file_path).split('\n\n')
|
265 |
+
|
266 |
+
for block in blocks:
|
267 |
+
if block.strip() != '':
|
268 |
+
lines = block.strip().split('\n')
|
269 |
+
index = lines[0]
|
270 |
+
time_line = lines[1].split(" --> ")
|
271 |
+
start, end = time_str_to_seconds(time_line[0], self.decimal_marker), time_str_to_seconds(time_line[1], self.decimal_marker)
|
272 |
+
sentence = ' '.join(lines[2:])
|
273 |
+
|
274 |
+
segments.append(Segment(
|
275 |
+
start=start,
|
276 |
+
end=end,
|
277 |
+
text=sentence
|
278 |
+
))
|
279 |
+
|
280 |
+
return segments
|
281 |
+
|
282 |
|
283 |
class WriteTSV(ResultWriter):
|
284 |
"""
|
|
|
314 |
def get_writer(
|
315 |
output_format: str, output_dir: str
|
316 |
) -> Callable[[dict, TextIO, dict], None]:
|
317 |
+
output_format = output_format.strip().lower().replace(".", "")
|
318 |
|
319 |
writers = {
|
320 |
"txt": WriteTXT,
|
|
|
341 |
def generate_file(
|
342 |
output_format: str, output_dir: str, result: Union[dict, List[Segment]], output_file_name: str, add_timestamp: bool = True,
|
343 |
) -> Tuple[str, str]:
|
344 |
+
output_format = output_format.strip().lower().replace(".", "")
|
345 |
+
|
346 |
+
if add_timestamp:
|
347 |
+
timestamp = datetime.now().strftime("%m%d%H%M%S")
|
348 |
+
output_file_name += timestamp
|
349 |
+
|
350 |
file_path = os.path.join(output_dir, f"{output_file_name}.{output_format}")
|
351 |
file_writer = get_writer(output_format=output_format, output_dir=output_dir)
|
352 |
+
file_writer(result=result, output_file_name=output_file_name)
|
353 |
content = read_file(file_path)
|
354 |
return content, file_path
|
355 |
|
356 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
357 |
def safe_filename(name):
|
358 |
INVALID_FILENAME_CHARS = r'[<>:"/\\|?*\x00-\x1f]'
|
359 |
safe_name = re.sub(INVALID_FILENAME_CHARS, '_', name)
|