input_ids list | chunk_idx uint32 0 62.5k | domain stringclasses 2
values | source_file stringclasses 12
values |
|---|---|---|---|
[
79448,
22614,
565,
220,
2955,
320,
34,
8,
220,
17,
15,
16,
18,
12,
17,
15,
16,
19,
350,
402,
8691,
31477,
198,
22614,
565,
220,
10103,
1212,
279,
8914,
1876,
11,
6079,
220,
17,
13,
15,
320,
1782,
330,
9827,
797,
565,
220,
498,
... | 0 | code | deduped_0043.jsonl |
[
488,
526,
7,
15,
13,
18,
22,
20,
23019,
9,
1054,
701,
29663,
16,
58,
16,
60,
488,
526,
7,
15,
13,
18,
22,
20,
23019,
32714,
85,
1171,
503,
4174,
17,
284,
320,
5173,
16,
58,
15,
60,
488,
526,
7,
15,
13,
17,
20,
9,
1054,
7... | 1 | code | deduped_0043.jsonl |
[
20095,
304,
656,
60502,
3394,
510,
310,
656,
7007,
62,
35782,
6085,
2561,
20095,
58,
16,
1125,
20095,
58,
17,
1125,
20095,
58,
18,
60,
10654,
20095,
58,
15,
9604,
262,
707,
4038,
1089,
960,
1193,
982,
286,
53705,
17,
2599,
284,
5297... | 2 | code | deduped_0043.jsonl |
[
340,
286,
5544,
17,
91668,
1193,
9140,
11,
2383,
11,
29663,
11,
656,
13990,
11,
220,
15,
13,
22,
345,
503,
656,
7961,
11,
656,
10932,
7927,
11,
5544,
17,
81353,
81941,
692,
262,
671,
4038,
5042,
10971,
3754,
304,
220,
18,
67,
198,... | 3 | code | deduped_0043.jsonl |
[
11,
1221,
2647,
198,
286,
770,
1855,
198,
286,
3190,
286,
2615,
6458,
11,
1196,
11,
15867,
284,
716,
455,
47396,
1193,
8223,
340,
286,
31004,
284,
48923,
13274,
28,
5211,
6458,
11,
1196,
30164,
11,
15867,
28,
23476,
340,
286,
1736,
... | 4 | code | deduped_0043.jsonl |
[
861,
21227,
59625,
16466,
198,
262,
44732,
262,
393,
7634,
2822,
31,
18559,
82,
1548,
33,
309,
340,
31,
4701,
90496,
50,
1183,
39626,
20546,
16322,
3075,
8467,
89079,
8097,
1843,
20180,
48654,
4739,
14618,
10,
4739,
4806,
14618,
10,
568... | 5 | code | deduped_0043.jsonl |
[
856,
20,
15,
24,
198,
1499,
86837,
860,
1370,
8470,
62465,
1159,
80733,
198,
1499,
86837,
860,
1370,
8470,
10367,
35209,
1159,
50257,
198,
1499,
86837,
1993,
20,
15,
24,
1159,
48766,
271,
1499,
342,
634,
14947,
1159,
66065,
957,
198,
... | 6 | code | deduped_0043.jsonl |
[
67753,
78,
1889,
3827,
96686,
8749,
48316,
691,
1245,
16,
16,
18,
11324,
9401,
262,
11456,
1066,
262,
671,
1140,
315,
821,
3542,
311,
7089,
198,
262,
44909,
284,
364,
44,
16,
16,
18,
4138,
32467,
11219,
1248,
262,
14448,
3136,
7791,
... | 7 | code | deduped_0043.jsonl |
[
286,
470,
320,
872,
34,
5584,
5879,
11,
5708,
34,
5584,
5879,
692,
262,
707,
738,
15960,
1193,
11,
1882,
2227,
982,
286,
421,
656,
556,
10412,
961,
2240,
510,
310,
656,
8659,
284,
1882,
2227,
670,
320,
721,
556,
10412,
340,
310,
4... | 8 | code | deduped_0043.jsonl |
[
3239,
58,
10033,
47993,
1457,
8696,
63824,
4189,
2533,
414,
3811,
284,
3489,
90,
59545,
15947,
1819,
16097,
6228,
338,
13868,
1778,
1078,
1269,
92,
314,
1631,
1269,
92,
6228,
338,
10545,
90,
1192,
8955,
92,
608,
314,
83192,
51084,
92,
... | 9 | code | deduped_0043.jsonl |
End of preview. Expand in Data Studio
Naime Corpus v1
多领域通用预训练语料库，使用 Qwen3-8B tokenizer 分词。
数据规格
| 指标 | 数值 |
|---|---|
| 总 Tokens | ~281亿 (28.1B) |
| 总文档数 | 38,186,704 |
| 序列长度 | 4096 |
| 格式 | Parquet (ZSTD) |
| 分片数 | 1736 |
| 总大小 | ~42 GB |
| 分词器 | Qwen/Qwen3-8B |
领域分布
| 领域 | 文档数 | 占比 |
|---|---|---|
| general_en | 21,608,669 | 56.6% |
| general_zh | 10,762,743 | 28.2% |
| wiki_en | 3,983,641 | 10.4% |
| wiki_zh | 844,764 | 2.2% |
| math | 628,510 | 1.6% |
| code | 358,377 | 0.9% |
Schema
input_ids: List[uint32] (length 4096)
chunk_idx: uint32
domain: string
source_file: string
使用方式
from datasets import load_dataset
ds = load_dataset("Leonharper/naime-corpus-v1", split="train")
en = ds.filter(lambda x: x["domain"] == "general_en")
- Downloads last month
- 1,854