Upload 2 files

meta.py

@@ -0,0 +1,233 @@
+lang2shard_cnt = {
+  'aa000': 2,
+  'ab000': 2,
+  'af000': 2,
+  'ak000': 2,
+  'am000': 9,
+  'ar000': 154,
+  'as000': 2,
+  'ay000': 2,
+  'az000': 4,
+  'ba000': 2,
+  'be000': 7,
+  'bg000': 12,
+  'bh000': 2,
+  'bi000': 2,
+  'bm000': 2,
+  'bn000': 92,
+  'bo000': 2,
+  'br000': 2,
+  'bs000': 2,
+  'ca000': 10,
+  'co000': 2,
+  'cr000': 2,
+  'cs000': 24,
+  'cy000': 2,
+  'da000': 6,
+  'de000': 369,
+  'de100': 500,
+  'de101': 500,
+  'de102': 114,
+  'dz000': 2,
+  'ee000': 2,
+  'el000': 18,
+  'en000': 500,
+  'en001': 500,
+  'en002': 500,
+  'en003': 500,
+  'en004': 500,
+  'en005': 500,
+  'en006': 500,
+  'en007': 437,
+  'en100': 500,
+  'en101': 500,
+  'en102': 500,
+  'en103': 500,
+  'en104': 500,
+  'en105': 500,
+  'en106': 500,
+  'en107': 500,
+  'en108': 500,
+  'en109': 500,
+  'en110': 500,
+  'en111': 500,
+  'en112': 500,
+  'en113': 500,
+  'en114': 500,
+  'en115': 500,
+  'en116': 500,
+  'en117': 500,
+  'en118': 500,
+  'en119': 500,
+  'en120': 500,
+  'en121': 500,
+  'en122': 500,
+  'en123': 500,
+  'en124': 500,
+  'en125': 500,
+  'en126': 500,
+  'en127': 500,
+  'en128': 500,
+  'en129': 62,
+  'eo000': 4,
+  'es000': 483,
+  'es100': 500,
+  'es101': 500,
+  'es102': 500,
+  'es103': 500,
+  'es104': 500,
+  'es105': 500,
+  'es106': 500,
+  'es107': 500,
+  'es108': 201,
+  'et000': 2,
+  'eu000': 4,
+  'fa000': 12,
+  'ff000': 2,
+  'fi000': 28,
+  'fj000': 2,
+  'fo000': 2,
+  'fr000': 315,
+  'fr100': 500,
+  'fr101': 500,
+  'fr102': 500,
+  'fr103': 401,
+  'fy000': 1,
+  'ga000': 2,
+  'gd000': 2,
+  'gl000': 3,
+  'gn000': 2,
+  'gu000': 8,
+  'ha000': 4,
+  'hi000': 182,
+  'hi100': 7,
+  'ho000': 2,
+  'hr000': 5,
+  'ht000': 3,
+  'hu000': 32,
+  'hy000': 3,
+  'ia000': 2,
+  'id000': 493,
+  'id100': 500,
+  'id101': 419,
+  'ie000': 2,
+  'ig000': 2,
+  'ik000': 2,
+  'is000': 2,
+  'it000': 185,
+  'it100': 500,
+  'it101': 432,
+  'iu000': 2,
+  'iw000': 21,
+  'ja000': 211,
+  'ja100': 303,
+  'jv000': 2,
+  'ka000': 4,
+  'ki000': 1,
+  'kk000': 6,
+  'kl000': 2,
+  'km000': 10,
+  'kn000': 7,
+  'ko000': 391,
+  'ko100': 500,
+  'ko101': 500,
+  'ko102': 500,
+  'ko103': 287,
+  'ks000': 2,
+  'ku000': 2,
+  'ky000': 4,
+  'la000': 2,
+  'lb000': 2,
+  'lg000': 2,
+  'ln000': 2,
+  'lo000': 2,
+  'lt000': 4,
+  'lv000': 2,
+  'mg000': 2,
+  'mi000': 2,
+  'mk000': 2,
+  'ml000': 12,
+  'mn000': 2,
+  'mr000': 18,
+  'ms000': 8,
+  'mt000': 0,
+  'my000': 2,
+  'na000': 2,
+  'nd000': 1,
+  'ne000': 6,
+  'nl000': 52,
+  'nl100': 263,
+  'no000': 17,
+  'nv000': 2,
+  'oc000': 2,
+  'om000': 2,
+  'or000': 3,
+  'pa000': 5,
+  'pl000': 140,
+  'ps000': 2,
+  'pt000': 202,
+  'pt100': 500,
+  'pt101': 500,
+  'pt102': 500,
+  'pt103': 382,
+  'qu000': 2,
+  'rm000': 2,
+  'rn000': 2,
+  'ro000': 18,
+  'ru000': 500,
+  'ru001': 287,
+  'ru100': 500,
+  'ru101': 500,
+  'ru102': 500,
+  'ru103': 500,
+  'ru104': 500,
+  'ru105': 500,
+  'ru106': 439,
+  'rw000': 2,
+  'sa000': 2,
+  'sc000': 2,
+  'sd000': 2,
+  'sg000': 1,
+  'sh000': 1,
+  'si000': 8,
+  'sk000': 6,
+  'sl000': 4,
+  'sm000': 2,
+  'sn000': 2,
+  'so000': 4,
+  'sq000': 2,
+  'sr000': 4,
+  'st000': 2,
+  'su000': 2,
+  'sv000': 17,
+  'sw000': 4,
+  'ta000': 40,
+  'te000': 14,
+  'tg000': 2,
+  'th000': 113,
+  'th100': 2,
+  'ti000': 2,
+  'tk000': 2,
+  'tn000': 2,
+  'to000': 2,
+  'tr000': 155,
+  'tr100': 440,
+  'ts000': 1,
+  'tt000': 2,
+  'ug000': 2,
+  'uk000': 63,
+  'uk100': 56,
+  'ur000': 35,
+  'uz000': 8,
+  've000': 2,
+  'vi000': 465,
+  'vi100': 500,
+  'vi101': 472,
+  'vo000': 2,
+  'wo000': 2,
+  'xh000': 2,
+  'yi000': 2,
+  'yo000': 2,
+  'zh000': 42,
+  'zu000': 2,
+}

yodas2.py

@@ -0,0 +1,206 @@
+import os
+from collections import OrderedDict
+from pathlib import Path
+import datasets
+import os
+from .meta import lang2shard_cnt
+import json
+
+
+class Yodas2Config(datasets.BuilderConfig):
+    """BuilderConfig for Yodas2."""
+
+    def __init__(self, lang, version, **kwargs):
+        self.language = lang
+        self.base_data_path = f"data/{lang}"
+
+        description = (
+            f"Youtube speech to text dataset in {self.language}."
+        )
+        super(Yodas2Config, self).__init__(
+            name=lang,
+            version=datasets.Version(version),
+            description=description,
+            **kwargs,
+        )
+
+
+DEFAULT_CONFIG_NAME = "all"
+LANGS = list(lang2shard_cnt.keys())
+VERSION = "1.0.0"
+
+class Yodas2(datasets.GeneratorBasedBuilder):
+    """YodasSample dataset."""
+
+    BUILDER_CONFIGS = [
+        Yodas2Config(lang, version=VERSION) for lang in LANGS
+    ]
+
+    VERSION = datasets.Version("1.0.1")
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description="Yodas Sample",
+            features=datasets.Features(
+                OrderedDict(
+                    [
+                        ("id", datasets.Value("string")),
+                        ("video_id", datasets.Value("string")),
+                        ('duration', datasets.Value('float')),
+                        ("audio", datasets.Audio(sampling_rate=24_000)),
+                        ("utterances", datasets.Sequence(feature={'utt_id': datasets.Value(dtype='string'),
+                                                                     'text': datasets.Value(dtype='string'),
+                                                                     'start': datasets.Value(dtype='float'),
+                                                                     'end': datasets.Value(dtype='float')}))
+                    ]
+                )
+            ),
+            supervised_keys=None,
+            homepage="",  # TODO
+            citation="",  # TODO
+        )
+
+
+    def _split_generators(self, dl_manager):
+        """Returns SplitGenerators."""
+        # TODO
+
+
+        total_cnt = lang2shard_cnt[self.config.name]
+
+        idx_lst = [f"{i:08d}" for i in range(total_cnt)]
+        audio_tar_files = dl_manager.download([f"{self.config.base_data_path}/audio/{i:08d}.tar.gz" for i in range(total_cnt)])
+        text_files = dl_manager.download([f"{self.config.base_data_path}/text/{i:08d}.json" for i in range(total_cnt)])
+        duration_files = dl_manager.download([f"{self.config.base_data_path}/duration/{i:08d}.txt" for i in range(total_cnt)])
+
+        if dl_manager.is_streaming:
+            audio_archives = [dl_manager.iter_archive(audio_tar_file) for audio_tar_file in audio_tar_files]
+            text_archives = [dl_manager.extract(text_file) for text_file in text_files]
+            duration_archives = [dl_manager.extract(duration_file) for duration_file in duration_files]
+
+        else:
+            print("extracting audio ...")
+            print("audio tarfiles... ")
+            print(audio_tar_files)
+            extracted_audio_archives = dl_manager.extract(audio_tar_files)
+            print("extracted archives...")
+            print(extracted_audio_archives)
+
+            audio_archives = []
+            text_archives = []
+            duration_archives = []
+            for idx, audio_tar_file, extracted_dir, text_file, duration_file in zip(idx_lst, audio_tar_files, extracted_audio_archives, text_files, duration_files):
+                audio_archives.append(extracted_dir)
+                text_archives.append(text_file)
+                duration_archives.append(duration_file)
+
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "is_streaming": dl_manager.is_streaming,
+                    "audio_archives": audio_archives,
+                    'text_archives': text_archives,
+                    'duration_archives': duration_archives,
+                },
+            ),
+        ]
+
+    def _generate_examples(self, is_streaming, audio_archives, text_archives, duration_archives):
+        """Yields examples."""
+
+        global_id = 0
+
+        if is_streaming:
+            for tar_file, text_file, duration_file in zip(audio_archives, text_archives, duration_archives):
+
+                # video to text
+                video2text = {}
+
+                json_obj_lst = json.loads(open(text_file, 'r').read())
+                for json_obj in json_obj_lst:
+                    video_id = json_obj['audio_id']
+                    video2text[video_id] = []
+
+                    for k,v in sorted(json_obj['text'].items()):
+                        fields = k.split('-')
+                        start_timestamp = float(fields[-2]) / 100
+                        end_timestamp = float(fields[-1]) / 100
+                        video2text[video_id].append({'utt_id': k, 'text': v, 'start': start_timestamp, 'end': end_timestamp})
+
+
+                # video to duration
+                video2duration = {}
+                with open(duration_file) as f:
+                    for id_, row in enumerate(f):
+                        fields = row.strip().split()
+                        video_id = fields[0]
+                        duration = float(fields[1])
+                        video2duration[video_id] = duration
+
+
+                for path, audio_f in tar_file:
+
+                    path = Path(path)
+                    video_id = path.stem
+
+                    if video_id in video2text and video_id in video2duration:
+
+                        result = {
+                            'id': global_id,
+                            'video_id': video_id,
+                            'audio': {"path": None, "bytes": audio_f.read()},
+                            'duration': video2duration[video_id],
+                            'utterances': video2text[video_id]
+                        }
+
+                        yield global_id, result
+                        global_id += 1
+        else:
+            for extracted_dir, text_file, duration_file in zip(audio_archives, text_archives, duration_archives):
+
+                print('extracted_dir ', extracted_dir)
+
+                print('actual extracted dir', extracted_dir)
+
+                # video to text
+                video2text = {}
+                json_obj_lst = json.loads(open(text_file, 'r').read())
+                for json_obj in json_obj_lst:
+                    video_id = json_obj['audio_id']
+                    video2text[video_id] = []
+
+                    for k,v in sorted(json_obj['text'].items()):
+                        fields = k.split('-')
+                        start_timestamp = float(fields[-2]) / 100
+                        end_timestamp = float(fields[-1]) / 100
+                        video2text[video_id].append({'utt_id': k, 'text': v, 'start': start_timestamp, 'end': end_timestamp})
+
+
+                # video to duration
+                video2duration = {}
+                with open(duration_file) as f:
+                    for id_, row in enumerate(f):
+                        fields = row.strip().split()
+                        video_id = fields[0]
+                        duration = float(fields[1])
+                        video2duration[video_id] = duration
+
+
+                for audio_file in list(Path(extracted_dir).glob('*')):
+
+                    video_id = audio_file.stem
+
+                    if video_id in video2text and video_id in video2duration:
+
+                        result = {
+                            'id': global_id,
+                            'video_id': video_id,
+                            'duration': video2duration[video_id],
+                            'audio': {"path": str(audio_file.absolute()), "bytes": open(audio_file, 'rb').read()},
+                            'utterances': video2text[video_id]
+                        }
+
+                        yield global_id, result
+                        global_id += 1