Datasets:
Repack: pair .wav+.txt inside the same tar (WebDataset layout) and drop standalone texts_part_*.tar
Browse files- README.md +18 -24
- texts_part_001.tar +0 -3
- texts_part_002.tar +0 -3
- texts_part_003.tar +0 -3
- wavs_part_001.tar +2 -2
- wavs_part_002.tar +2 -2
- wavs_part_003.tar +2 -2
README.md
CHANGED
|
@@ -53,30 +53,27 @@ ambiguity and dramatically improves prosody.
|
|
| 53 |
|
| 54 |
```
|
| 55 |
metadata.csv # 22,200 lines, pipe-delimited: filename|text_with_stress
|
| 56 |
-
wavs_part_001.tar #
|
| 57 |
-
wavs_part_002.tar #
|
| 58 |
-
wavs_part_003.tar #
|
| 59 |
-
texts_part_001.tar # TXT files 000000…007399 (≈7 MB)
|
| 60 |
-
texts_part_002.tar # TXT files 007400…014799 (≈7 MB)
|
| 61 |
-
texts_part_003.tar # TXT files 014800…022199 (≈7 MB)
|
| 62 |
```
|
| 63 |
|
| 64 |
The data is laid out in the **WebDataset** convention: each sample consists of
|
| 65 |
a `.wav` and a `.txt` sharing the same key (`000123_RUSLAN.wav` ↔
|
| 66 |
-
`000123_RUSLAN.txt`)
|
|
|
|
| 67 |
|
| 68 |
`metadata.csv` is provided as a convenience for users who only want the text;
|
| 69 |
it is the exact concatenation of all `.txt` files (`filename|text` per line).
|
| 70 |
|
| 71 |
The audio is split into three uncompressed `.tar` shards instead of being
|
| 72 |
uploaded as 22,200 individual files (HuggingFace rate-limits per-file commits
|
| 73 |
-
on free tier). After download, extract every shard to get flat
|
| 74 |
-
`
|
| 75 |
|
| 76 |
```bash
|
| 77 |
-
mkdir -p
|
| 78 |
-
for f in wavs_part_*.tar;
|
| 79 |
-
for f in texts_part_*.tar; do tar -xf "$f" -C texts; done
|
| 80 |
```
|
| 81 |
|
| 82 |
Total: 22,200 mono 22 kHz WAV files (≈31 hours of speech) + matching `.txt`
|
|
@@ -166,29 +163,26 @@ When using this dataset please cite the original RUSLAN paper:
|
|
| 166 |
|
| 167 |
```
|
| 168 |
metadata.csv # 22 200 строк, формат filename|текст_с_ударениями (разделитель |)
|
| 169 |
-
wavs_part_001.tar #
|
| 170 |
-
wavs_part_002.tar #
|
| 171 |
-
wavs_part_003.tar #
|
| 172 |
-
texts_part_001.tar # TXT-файлы 000000…007399 (≈7 МБ)
|
| 173 |
-
texts_part_002.tar # TXT-файлы 007400…014799 (≈7 МБ)
|
| 174 |
-
texts_part_003.tar # TXT-файлы 014800…022199 (≈7 МБ)
|
| 175 |
```
|
| 176 |
|
| 177 |
Данные оформлены по соглашению **WebDataset**: один сэмпл состоит из `.wav` и
|
| 178 |
-
`.txt` с одинаковым ключом (`000123_RUSLAN.wav` ↔ `000123_RUSLAN.txt`)
|
| 179 |
-
|
|
|
|
| 180 |
|
| 181 |
`metadata.csv` приложен как удобство для тех, кому нужен только текст; он
|
| 182 |
содержит ровно те же строки, что и `.txt` файлы (`filename|текст` в строку).
|
| 183 |
|
| 184 |
Аудио сложено в три tar-шарда (без сжатия) вместо 22 200 отдельных файлов —
|
| 185 |
бесплатный тариф HuggingFace ограничивает коммиты пофайлово. После скачивания
|
| 186 |
-
распакуйте все шарды:
|
| 187 |
|
| 188 |
```bash
|
| 189 |
-
mkdir -p
|
| 190 |
-
for f in wavs_part_*.tar;
|
| 191 |
-
for f in texts_part_*.tar; do tar -xf "$f" -C texts; done
|
| 192 |
```
|
| 193 |
|
| 194 |
Итого: 22 200 моно WAV-файлов 22 кГц (≈31 час речи) + соответствующие `.txt`
|
|
|
|
| 53 |
|
| 54 |
```
|
| 55 |
metadata.csv # 22,200 lines, pipe-delimited: filename|text_with_stress
|
| 56 |
+
wavs_part_001.tar # samples 000000…007399 (≈2.1 GB) — wav + paired txt
|
| 57 |
+
wavs_part_002.tar # samples 007400…014799 (≈3.2 GB) — wav + paired txt
|
| 58 |
+
wavs_part_003.tar # samples 014800…022199 (≈4.2 GB) — wav + paired txt
|
|
|
|
|
|
|
|
|
|
| 59 |
```
|
| 60 |
|
| 61 |
The data is laid out in the **WebDataset** convention: each sample consists of
|
| 62 |
a `.wav` and a `.txt` sharing the same key (`000123_RUSLAN.wav` ↔
|
| 63 |
+
`000123_RUSLAN.txt`), both stored together inside the same `.tar` shard. The
|
| 64 |
+
HuggingFace dataset viewer pairs them automatically.
|
| 65 |
|
| 66 |
`metadata.csv` is provided as a convenience for users who only want the text;
|
| 67 |
it is the exact concatenation of all `.txt` files (`filename|text` per line).
|
| 68 |
|
| 69 |
The audio is split into three uncompressed `.tar` shards instead of being
|
| 70 |
uploaded as 22,200 individual files (HuggingFace rate-limits per-file commits
|
| 71 |
+
on free tier). After download, extract every shard to get a flat folder with
|
| 72 |
+
matched `.wav`/`.txt` pairs:
|
| 73 |
|
| 74 |
```bash
|
| 75 |
+
mkdir -p data
|
| 76 |
+
for f in wavs_part_*.tar; do tar -xf "$f" -C data; done
|
|
|
|
| 77 |
```
|
| 78 |
|
| 79 |
Total: 22,200 mono 22 kHz WAV files (≈31 hours of speech) + matching `.txt`
|
|
|
|
| 163 |
|
| 164 |
```
|
| 165 |
metadata.csv # 22 200 строк, формат filename|текст_с_ударениями (разделитель |)
|
| 166 |
+
wavs_part_001.tar # сэмплы 000000…007399 (≈2,1 ГБ) — wav + парный txt
|
| 167 |
+
wavs_part_002.tar # сэмплы 007400…014799 (≈3,2 ГБ) — wav + парный txt
|
| 168 |
+
wavs_part_003.tar # сэмплы 014800…022199 (≈4,2 ГБ) — wav + парный txt
|
|
|
|
|
|
|
|
|
|
| 169 |
```
|
| 170 |
|
| 171 |
Данные оформлены по соглашению **WebDataset**: один сэмпл состоит из `.wav` и
|
| 172 |
+
`.txt` с одинаковым ключом (`000123_RUSLAN.wav` ↔ `000123_RUSLAN.txt`), оба
|
| 173 |
+
лежат **внутри одного и того же tar-шарда**. HF dataset viewer склеивает их
|
| 174 |
+
автоматически и показывает текст рядом с аудио.
|
| 175 |
|
| 176 |
`metadata.csv` приложен как удобство для тех, кому нужен только текст; он
|
| 177 |
содержит ровно те же строки, что и `.txt` файлы (`filename|текст` в строку).
|
| 178 |
|
| 179 |
Аудио сложено в три tar-шарда (без сжатия) вместо 22 200 отдельных файлов —
|
| 180 |
бесплатный тариф HuggingFace ограничивает коммиты пофайлово. После скачивания
|
| 181 |
+
распакуйте все шарды в общую папку:
|
| 182 |
|
| 183 |
```bash
|
| 184 |
+
mkdir -p data
|
| 185 |
+
for f in wavs_part_*.tar; do tar -xf "$f" -C data; done
|
|
|
|
| 186 |
```
|
| 187 |
|
| 188 |
Итого: 22 200 моно WAV-файлов 22 кГц (≈31 час речи) + соответствующие `.txt`
|
texts_part_001.tar
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:150f95ec294cdccfc8ccd57517cd7161699c63e3643bae42af441a02ebe1e455
|
| 3 |
-
size 7587840
|
|
|
|
|
|
|
|
|
|
|
|
texts_part_002.tar
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:1eafb9e21d90ace0cf90cb0b2b469e516d6c937df8b5923591b49d5fe2af7675
|
| 3 |
-
size 7587840
|
|
|
|
|
|
|
|
|
|
|
|
texts_part_003.tar
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:c55517a7bdafbbb5100720d0f6f6e1f2fd98cb6488057782bd9d9e9f5f006b17
|
| 3 |
-
size 7690240
|
|
|
|
|
|
|
|
|
|
|
|
wavs_part_001.tar
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c6b66a4f05f1f957a557fa102c4606b848213147456c0584f7827fe029fa8182
|
| 3 |
+
size 2207559680
|
wavs_part_002.tar
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:88f03753b380b024fbb96e4ffe159180ed9d89d443f28230d4240a79e876f5cb
|
| 3 |
+
size 3411425280
|
wavs_part_003.tar
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77e177240a0dce9ddef056b34abbf4785f591e10b27c4e89e492b0a770d53169
|
| 3 |
+
size 4439224320
|