roro128 commited on
Commit
8a31ec4
·
verified ·
1 Parent(s): 345de23

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +28 -40
README.md CHANGED
@@ -36,63 +36,51 @@ size_categories:
36
  - 100G<n<1T
37
  ---
38
 
39
- # FLEURS-FLAC: Losslessly FLAC-Compressed Google FLEURS Dataset
40
 
41
- `roro128/fleurs-flac`은 Google의 **FLEURS (Few-shot Learning Evaluation of Universal Representations of Speech)** 오디오 데이터셋(102개 언어)을 원본 음질 손상 없이 **FLAC (PCM24 무손실) 포맷으로 압축**하고 **500MB Parquet 샤드** 단위로 최적화하여 배포하는 데이터셋입니다.
42
 
43
- ## 🚀 주요 특징 (Key Features)
44
 
45
- - **FLAC 무손실 압축**: 원본 오디오 샘플 해시 및 PCM 무손실 검증 완료.
46
- - **고속 디코딩 및 스트리밍**: 학습 및 추론 시 오디오 해제/읽기 속도를 극대화.
47
- - **500MB Parquet Sharding**: 메모리 및 네트워크 대역폭에 최적화된 샤드 단위 관리.
48
- - **102개 전체 언어 완벽 지원**: `train`, `validation`, `test` 분할 완벽 수록.
49
 
50
- ## 💻 사용법 (Usage)
 
 
 
51
 
52
- Hugging Face `datasets` 라이브러리를 사용하여 간편하게 로드할 수 있습니다:
53
 
54
  ```python
55
- from datasets import load_dataset
56
 
57
- # 특정 언어 로드 (예: 한국어 'ko_kr')
58
  dataset = load_dataset("roro128/fleurs-flac", "ko_kr")
59
 
60
- # 데이터셋 분할 접근
61
- train_data = dataset["train"]
62
- print(train_data[0])
63
- ```
64
-
65
- ### 오디오 디코딩 사용예시
66
- ```python
67
- from datasets import Audio, load_dataset
68
-
69
- dataset = load_dataset("roro128/fleurs-flac", "ko_kr")
70
-
71
- # 오디오 기능 디코딩 설정
72
  dataset = dataset.cast_column("audio", Audio())
73
- sample = dataset["train"][0]
74
 
75
- print("Sample Rate:", sample["audio"]["sampling_rate"])
76
- print("Audio Array:", sample["audio"]["array"])
 
77
  ```
78
 
79
- ## 📊 데이터셋 구조 (Dataset Structure)
80
 
81
- | Column | Type | Description |
82
- | :--- | :--- | :--- |
83
- | `id` | `int64` | 샘플 식별 고유 ID |
84
- | `num_samples` | `int64` | PCM 샘플 개수 |
85
- | `raw_transcription` | `string` | 원본 전사 텍스트 |
86
- | `transcription` | `string` | 정제된 전사 텍스트 |
87
- | `gender` | `int64` | 성별 (0: MALE, 1: FEMALE) |
88
- | `path` | `string` | 오디오 상대 경로 |
89
- | `language` | `string` | 언어 코드 (예: `af_za`, `ko_kr`) |
90
- | `audio` | `dict` | `{"bytes": binary, "path": string}` FLAC 오디오 데이터 |
91
 
92
- ## 🌐 지원 언어 목록 (102 Languages)
93
 
 
94
  `af_za`, `am_et`, `ar_eg`, `as_in`, `ast_es`, `az_az`, `be_by`, `bg_bg`, `bn_in`, `bs_ba`, `ca_es`, `ceb_ph`, `ckb_iq`, `cmn_hans_cn`, `cs_cz`, `cy_gb`, `da_dk`, `de_de`, `el_gr`, `en_us`, `es_419`, `et_ee`, `fa_ir`, `ff_sn`, `fi_fi`, `fil_ph`, `fr_fr`, `ga_ie`, `gl_es`, `gu_in`, `ha_ng`, `he_il`, `hi_in`, `hr_hr`, `hu_hu`, `hy_am`, `id_id`, `ig_ng`, `is_is`, `it_it`, `ja_jp`, `jv_id`, `ka_ge`, `kam_ke`, `kea_cv`, `kk_kz`, `km_kh`, `kn_in`, `ko_kr`, `ky_kg`, `lb_lu`, `lg_ug`, `ln_cd`, `lo_la`, `lt_lt`, `luo_ke`, `lv_lv`, `mi_nz`, `mk_mk`, `ml_in`, `mn_mn`, `mr_in`, `ms_my`, `mt_mt`, `my_mm`, `nb_no`, `ne_np`, `nl_nl`, `nso_za`, `ny_mw`, `oc_fr`, `om_et`, `or_in`, `pa_in`, `pl_pl`, `ps_af`, `pt_br`, `ro_ro`, `ru_ru`, `sd_in`, `sk_sk`, `sl_si`, `sn_zw`, `so_so`, `sq_al`, `sr_rs`, `sv_se`, `sw_ke`, `ta_in`, `te_in`, `tg_tj`, `th_th`, `tr_tr`, `uk_ua`, `ur_pk`, `uz_uz`, `vi_vn`, `wo_sn`, `xh_za`, `yo_ng`, `yue_hant_hk`, `zu_za`.
95
 
96
- ## 📜 라이선스 (License)
97
 
98
- 원본 데이터셋과 동일하게 **CC-BY-4.0** 라이선스를 따릅니다.
 
36
  - 100G<n<1T
37
  ---
38
 
39
+ # FLEURS-FLAC
40
 
41
+ A losslessly FLAC-compressed version of Google's FLEURS dataset covering 102 languages.
42
 
43
+ ## Overview
44
 
45
+ This repository contains the Google FLEURS dataset repackaged into Parquet shards with PCM24 FLAC-compressed audio binaries.
 
 
 
46
 
47
+ Key points:
48
+ - Audio streams are converted to FLAC (PCM24) with sample-level PCM verification against the source.
49
+ - Sharded into ~500MB Parquet files per split for efficient I/O and streaming.
50
+ - Covers all 102 languages from the original FLEURS benchmark across `train`, `validation`, and `test` splits.
51
 
52
+ ## Usage
53
 
54
  ```python
55
+ from datasets import load_dataset, Audio
56
 
57
+ # Load a specific language config (e.g. ko_kr)
58
  dataset = load_dataset("roro128/fleurs-flac", "ko_kr")
59
 
60
+ # Cast audio column for automatic decoding
 
 
 
 
 
 
 
 
 
 
 
61
  dataset = dataset.cast_column("audio", Audio())
 
62
 
63
+ sample = dataset["train"][0]
64
+ print(sample["audio"]["sampling_rate"])
65
+ print(sample["audio"]["array"])
66
  ```
67
 
68
+ ## Dataset Fields
69
 
70
+ - `id` (int64): Sample identifier.
71
+ - `num_samples` (int64): Number of PCM audio samples.
72
+ - `raw_transcription` (string): Raw unnormalized text transcription.
73
+ - `transcription` (string): Normalized text transcription.
74
+ - `gender` (int64): Speaker gender (0: MALE, 1: FEMALE).
75
+ - `path` (string): Audio file path.
76
+ - `language` (string): Language config code (e.g. `af_za`, `ko_kr`).
77
+ - `audio` (struct): Dictionary containing `bytes` (FLAC binary) and `path`.
 
 
78
 
79
+ ## Languages
80
 
81
+ Includes all 102 FLEURS languages:
82
  `af_za`, `am_et`, `ar_eg`, `as_in`, `ast_es`, `az_az`, `be_by`, `bg_bg`, `bn_in`, `bs_ba`, `ca_es`, `ceb_ph`, `ckb_iq`, `cmn_hans_cn`, `cs_cz`, `cy_gb`, `da_dk`, `de_de`, `el_gr`, `en_us`, `es_419`, `et_ee`, `fa_ir`, `ff_sn`, `fi_fi`, `fil_ph`, `fr_fr`, `ga_ie`, `gl_es`, `gu_in`, `ha_ng`, `he_il`, `hi_in`, `hr_hr`, `hu_hu`, `hy_am`, `id_id`, `ig_ng`, `is_is`, `it_it`, `ja_jp`, `jv_id`, `ka_ge`, `kam_ke`, `kea_cv`, `kk_kz`, `km_kh`, `kn_in`, `ko_kr`, `ky_kg`, `lb_lu`, `lg_ug`, `ln_cd`, `lo_la`, `lt_lt`, `luo_ke`, `lv_lv`, `mi_nz`, `mk_mk`, `ml_in`, `mn_mn`, `mr_in`, `ms_my`, `mt_mt`, `my_mm`, `nb_no`, `ne_np`, `nl_nl`, `nso_za`, `ny_mw`, `oc_fr`, `om_et`, `or_in`, `pa_in`, `pl_pl`, `ps_af`, `pt_br`, `ro_ro`, `ru_ru`, `sd_in`, `sk_sk`, `sl_si`, `sn_zw`, `so_so`, `sq_al`, `sr_rs`, `sv_se`, `sw_ke`, `ta_in`, `te_in`, `tg_tj`, `th_th`, `tr_tr`, `uk_ua`, `ur_pk`, `uz_uz`, `vi_vn`, `wo_sn`, `xh_za`, `yo_ng`, `yue_hant_hk`, `zu_za`.
83
 
84
+ ## License
85
 
86
+ CC-BY-4.0 (matching the original google/fleurs dataset license).