Upload dataset.py

dataset.py

@@ -1,48 +1,191 @@
-import os
+# coding=utf-8
+"""
+new_dataset_stt_audio dataset.
+
+This dataset consists of audio files stored in tar archives and transcript files in TSV format.
+The dataset structure is as follows:
+
+new_dataset_stt_audio/
+├── audio/
+│   └── uz/
+│       ├── train/
+│       │   └── train.tar
+│       ├── validation/
+│       │   └── validation.tar
+│       └── test/
+│           └── test.tar
+└── transcript/
+    └── uz/
+        ├── train/
+        │   └── train.tsv
+        ├── validation/
+        │   └── validation.tsv
+        └── test/
+             └── test.tsv
+
+Each transcript TSV file has columns:
+id, path, sentence, duration, age, gender, accents, locale.
+
+The audio field is loaded using a tar URI, allowing streaming from the tar archive.
+"""
+
 import csv
+import os
 import tarfile
+from typing import Iterator, Tuple
+
+import datasets
+
+_CITATION = """\
+@misc{yourcitation2023,
+  title={Your Dataset Title},
+  author={Your Name},
+  year={2023},
+  url={https://huggingface.co/datasets/Elyordev/new_dataset_stt_audio}
+}
+"""
+
+_DESCRIPTION = """\
+This dataset consists of audio files and corresponding transcripts for speech-to-text tasks.
+The audio files are stored in tar archives under the audio/uz folder for each split (train, validation, test),
+and the transcripts are stored as TSV files under transcript/uz for each split.
+The transcript TSV files have the following columns:
+id, path, sentence, duration, age, gender, accents, locale.
+The audio is loaded using a tar URI to enable streaming.
+"""
+
+_HOMEPAGE = "https://huggingface.co/datasets/Elyordev/new_dataset_stt_audio"
+_LICENSE = "MIT"
+
+class NewDatasetSTTAudioConfig(datasets.BuilderConfig):
+    """Builder config for new_dataset_stt_audio."""
+    def __init__(self, language="uz", **kwargs):
+        super(NewDatasetSTTAudioConfig, self).__init__(**kwargs)
+        self.language = language
+
+class NewDatasetSTTAudio(datasets.GeneratorBasedBuilder):
+    """New Dataset STT Audio builder."""
+    VERSION = datasets.Version("1.0.0")
+    BUILDER_CONFIGS = [
+        NewDatasetSTTAudioConfig(
+            name="default",
+            version=VERSION,
+            description="STT dataset with audio tar archives and transcript TSV files for Uzbek language",
+            language="uz",
+        ),
+    ]
+
+    def _info(self):
+        features = datasets.Features({
+            "id": datasets.Value("string"),
+            "path": datasets.Value("string"),
+            "sentence": datasets.Value("string"),
+            "duration": datasets.Value("float"),
+            "age": datasets.Value("string"),
+            "gender": datasets.Value("string"),
+            "accents": datasets.Value("string"),
+            "locale": datasets.Value("string"),
+            "audio": datasets.Audio(sampling_rate=16000),
+        })
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        """
+        Returns SplitGenerators.
+        Expects the dataset to be provided manually via the repository.
+        The manual_dir should contain the following structure:
+        
+        new_dataset_stt_audio/
+          audio/uz/{train, validation, test}/*.tar
+          transcript/uz/{train, validation, test}/*.tsv
+        """
+        manual_dir = dl_manager.manual_dir if dl_manager.manual_dir is not None else ""
+        language = self.config.language
+
+        splits = {
+            "train": {
+                "transcript": os.path.join(manual_dir, "transcript", language, "train", "train.tsv"),
+                "audio": os.path.join(manual_dir, "audio", language, "train", "train.tar")
+            },
+            "validation": {
+                "transcript": os.path.join(manual_dir, "transcript", language, "validation", "validation.tsv"),
+                "audio": os.path.join(manual_dir, "audio", language, "validation", "validation.tar")
+            },
+            "test": {
+                "transcript": os.path.join(manual_dir, "transcript", language, "test", "test.tsv"),
+                "audio": os.path.join(manual_dir, "audio", language, "test", "test.tar")
+            }
+        }
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "transcript_path": splits["train"]["transcript"],
+                    "audio_tar_path": splits["train"]["audio"],
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                gen_kwargs={
+                    "transcript_path": splits["validation"]["transcript"],
+                    "audio_tar_path": splits["validation"]["audio"],
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={
+                    "transcript_path": splits["test"]["transcript"],
+                    "audio_tar_path": splits["test"]["audio"],
+                },
+            ),
+        ]
+
+    def _generate_examples(self, transcript_path: str, audio_tar_path: str) -> Iterator[Tuple[str, dict]]:
+        """
+        Yields examples as (key, example) tuples.
+
+        Args:
+            transcript_path (str): Path to the transcript TSV file.
+            audio_tar_path (str): Path to the audio tar archive.
+        """
+        # 1. Read the transcript TSV file into a dictionary mapping file name to metadata.
+        metadata_map = {}
+        with open(transcript_path, encoding="utf-8") as f:
+            reader = csv.DictReader(f, delimiter="\t")
+            for row in reader:
+                file_name = row["path"].strip()
+                if not file_name.endswith(".mp3"):
+                    file_name += ".mp3"
+                metadata_map[file_name] = row
 
-def _generate_examples(self, prompts_path, audio_tar_path):
-    """
-    Yields examples as (key, example) tuples.
-
-    Args:
-        prompts_path (str): transcript/uz/<split>/<split>.tsv – metadata fayl yo'li.
-        audio_tar_path (str): audio/uz/<split>/<split>.tar – tar arxiv yo'li (ichida *.mp3).
-    """
-    # 1. TSV fayldan metadata ni o'qish
-    metadata_map = {}
-    with open(prompts_path, encoding="utf-8") as f:
-        reader = csv.DictReader(f, delimiter="\t")
-        for row in reader:
-            # "path" ustunida fayl nomi bo'lishi kutiladi, masalan: "H3H38EY38D8.mp3"
-            file_name = row["path"].strip()
-            if not file_name.endswith(".mp3"):
-                file_name += ".mp3"
-            metadata_map[file_name] = row
-
-    # 2. Tar arxiv ichidan audio fayllarni o'qish
-    id_ = 0
-    with tarfile.open(audio_tar_path, "r") as tar:
-        for member in tar.getmembers():
-            # Tar ichidagi fayl nomidan faqat fayl nomini olish
-            file_name = os.path.basename(member.name)
-            if file_name in metadata_map:
-                row = metadata_map[file_name]
-                audio_file = tar.extractfile(member)
-                if audio_file is None:
-                    continue  # Agar fayl ochilmasa, o'tkazib yuboramiz
-                audio_bytes = audio_file.read()
-                audio = {"path": file_name, "bytes": audio_bytes}
-                yield id_, {
-                    "id": row.get("id", file_name),
-                    "path": row.get("path", file_name),
-                    "sentence": row.get("sentence", ""),
-                    "duration": float(row.get("duration", 0.0)),
-                    "age": row.get("age", ""),
-                    "gender": row.get("gender", ""),
-                    "accents": row.get("accents", ""),
-                    "locale": row.get("locale", ""),
-                    "audio": audio,
-                }
-                id_ += 1
+        # 2. Create a base audio URI for streaming from the tar archive.
+        base_audio_uri = f"tar://{audio_tar_path}#"
+        
+        # 3. Open the tar archive and iterate through its members.
+        id_ = 0
+        with tarfile.open(audio_tar_path, "r") as tar:
+            for member in tar.getmembers():
+                file_name = os.path.basename(member.name)
+                if file_name in metadata_map:
+                    row = metadata_map[file_name]
+                    audio_uri = base_audio_uri + file_name
+                    yield str(id_), {
+                        "id": row["id"],
+                        "path": row["path"],
+                        "sentence": row["sentence"],
+                        "duration": float(row.get("duration", 0.0)),
+                        "age": row.get("age", ""),
+                        "gender": row.get("gender", ""),
+                        "accents": row.get("accents", ""),
+                        "locale": row.get("locale", ""),
+                        "audio": audio_uri,
+                    }
+                    id_ += 1