Datasets:

Infinity08
/

KAWK500M-Korean-SFT-v1

Tasks:

Text Generation

Modalities:

Formats:

Languages:

Size:

Libraries:

License:

Dataset card Data Studio Files Files and versions

Dataset Viewer

Auto-converted to Parquet Duplicate

Split (2)

train · 22.2k rows

input_ids list	loss_mask list	source stringclasses 4 values
[ 1, 483, 2650, 2650, 1751, 315, 17, 328, 4546, 9533, 283, 294, 266, 5651, 328, 1333, 1468, 273, 11528, 327, 263, 17, 17, 483, 2650, 2650, 263, 13531, 315, 17, 799, 1768, 5894, 1022, 791, 1002, 265, 1375, 4949, 604, 6164, 665, 1251, ...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 10373, 1052, 7391, 263, 278, 434, 296, 269, 263, 270, 268, 447, 272, 12884, 265, 420, 1052, 7391, 263, 270, 434, 296, 269, 263, 268, 283, 447, 272, 11817, 2472, 265, 420, 1052, 1606, 267, 3277, 2...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 298, 639, 286, 404, 898, 10160, 3156, 310, 4016, 452, 1594, 5522, 1829, 1208, 7734, 288, 13465, 1380, 281, 863, 3721, 264, 5360, 1124, 286, 16374, 2254, 299, 321, 265, 7823, 7387, 307, 780, 3557, 1...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 6795, 6019, 9256, 2160, 1944, 312, 279, 6787, 327, 1549, 2122, 880, 470, 7103, 427, 2912, 264, 301, 2843, 4743, 586, 880, 11951, 327, 263, 17, 17, 483, 2650, 2650, 263, 13531, 315, 17, 6947, 880, ...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 6273, 9232, 12254, 271, 1335, 7889, 12455, 1429, 602, 269, 1787, 1619, 442, 1760, 408, 264, 1320, 9340, 2440, 291, 12455, 4875, 267, 3115, 3259, 1309, 7182, 375, 729, 2310, 264, 2357, 5442, 282, 1731...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 424, 3320, 3646, 272, 455, 709, 728, 18313, 327, 4516, 301, 3646, 272, 455, 2682, 312, 4091, 4501, 327, 263, 17, 17, 483, 2650, 2650, 263, 13531, 315, 17, 13990, 452, 2378, 1704, 1478, 8766, 9261, ...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 2538, 1626, 282, 421, 6801, 8728, 557, 274, 898, 1430, 272, 339, 6256, 269, 1312, 1314, 264, 2149, 10058, 5646, 291, 5538, 7307, 339, 3674, 275, 2540, 1314, 264, 263, 17, 17, 483, 2650, 2650, 263, ...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 14862, 1692, 685, 6502, 6194, 1556, 272, 339, 18518, 265, 931, 914, 1022, 278, 294, 294, 5969, 267, 2678, 347, 265, 461, 446, 710, 4713, 272, 10122, 2179, 5892, 1567, 10923, 2665, 3840, 275, 1312, ...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 270, 266, 270, 270, 308, 263, 294, 322, 263, 268, 294, 296, 265, 985, 316, 1045, 14419, 6032, 2198, 3201, 5273, 4058, 1118, 269, 2644, 1658, 5786, 13992, 267, 5621, 272, 1970, 5469, 348, 14368, 202...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 5772, 263, 17900, 2663, 276, 545, 644, 979, 336, 6630, 8101, 2599, 8859, 523, 362, 265, 7428, 267, 935, 802, 286, 11820, 288, 707, 998, 395, 3000, 282, 267, 4844, 3191, 269, 369, 562, 3275, 8179, ...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 14188, 452, 926, 976, 377, 388, 267, 1033, 263, 6635, 452, 4041, 276, 11528, 327, 263, 17, 17, 483, 2650, 2650, 263, 13531, 315, 17, 7469, 926, 976, 731, 489, 1033, 263, 6635, 452, 705, 4041, 271...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 9712, 557, 267, 3774, 4183, 1743, 333, 284, 18983, 277, 273, 263, 268, 268, 265, 266, 266, 266, 768, 295, 268, 2161, 295, 268, 1378, 265, 3992, 557, 275, 9787, 300, 563, 4410, 2981, 267, 15684, 2...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 443, 414, 276, 301, 5951, 2073, 7807, 267, 16784, 5773, 337, 264, 1890, 4755, 273, 1792, 263, 280, 452, 263, 272, 1055, 265, 2728, 1215, 609, 651, 699, 5965, 630, 576, 611, 265, 1307, 609, 4247, ...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 271, 2025, 515, 3057, 1146, 312, 279, 4501, 265, 3483, 3840, 3846, 274, 1086, 6558, 327, 263, 17, 17, 483, 2650, 2650, 263, 13531, 315, 17, 2572, 1005, 315, 17, 365, 526, 315, 17, 1201, 739, 1726...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1...	distilqwen
[ 1, 483, 2650, 2650, 1751, 315, 17, 379, 298, 481, 10769, 267, 3564, 263, 272, 434, 273, 2890, 263, 272, 434, 263, 10674, 463, 4134, 5091, 12174, 6135, 269, 3898, 3709, 307, 739, 4670, 3544, 1497, 300, 1829, 7005, 304, 264, 263, 272,...	[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...	distilqwen

End of preview. Expand in Data Studio

KAWK500M Korean SFT v1

한국어 전용 500M급 causal LM을 위한 assistant-only SFT 데이터입니다. 기존 Keural 대화/RAG 데이터는 사용하지 않았습니다.

고정된 원본

distilqwen: lcw99/DistilQwen_100k_korean @ 7324954c71af8f0974654e5c7e211f5b3fad5a96 (apache-2.0)
writing: coastral/korean-writing-style-instruct @ 9f31462314f07847534fd13f7406c330ab15d156 (apache-2.0)
carrot: CarrotAI/ko-instruction-dataset @ 5c0e2c0180b50400e401dd0b296043f18fc6cb3f (apache-2.0)
aya-human-ko: CohereLabs/aya_dataset @ f9ea04583f02a8f86404ff6c58bf75fe637df8a2 (apache-2.0)

처리

한국어 열만 사용하고 NFC 정규화
HTML 정리 및 이메일·전화번호·주민등록번호 형태 마스킹
명시적인 코드, LaTeX/수식 중심 샘플, 공개된 사고과정 태그 제거
한국어 비율 및 반복·기호·목록 품질 필터
모든 원본 사이 BLAKE2b 정확 중복 제거
KAWK500M 32K 토크나이저로 2,048토큰 고정 패킹
사용자·시스템 토큰은 loss_mask=0, 답변 및 EOS만 loss_mask=1
문서 해시 기반 고정 validation 분리

규모

train: 22,155 records
validation: 300 records
sequence length: 2,048
supervised assistant tokens: 34,174,461

각 행에는 input_ids, loss_mask, source가 있습니다. RunPod에서는 load_dataset(..., streaming=True)로 읽을 수 있습니다. 원본별 세부 조건과 제한도 함께 확인해야 합니다.

Downloads last month: 77

Number of rows:

Total file size:

Models trained or fine-tuned on Infinity08/KAWK500M-Korean-SFT-v1

Infinity08/KAWK-500M-Korean-Instruct-v1

Text Generation • 0.5B • Updated 2 days ago • 61