oz1115 commited on
Commit
467fc4d
·
verified ·
1 Parent(s): 1babed6

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +78 -0
  2. metadata.json +7 -0
  3. train.pkl +3 -0
  4. val.pkl +3 -0
README.md ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+
4
+ ko
5
+ license: mit
6
+ task_categories:
7
+ text-generation
8
+ size_categories:
9
+ 1K<n<10K
10
+
11
+
12
+ korean-pretraining-corpus
13
+ 한국어 사전학습 코퍼스
14
+ Dataset Description
15
+ 이 데이터셋은 한국어 Foundation Model 학습에 사용된 전처리된 코퍼스입니다.
16
+ Dataset Summary
17
+
18
+ Language: Korean (한국어)
19
+ License: MIT
20
+ Source: Korean Wikipedia and public Korean text
21
+ Preprocessing: Tokenized and chunked into sequences
22
+
23
+ Data Fields
24
+
25
+ input_ids: 토큰화된 정수 시퀀스 (List[int])
26
+
27
+ Data Splits
28
+
29
+ Train: ~95%
30
+ Validation: ~5%
31
+
32
+ Dataset Creation
33
+ Source Data
34
+ 한국어 위키피디아 및 공개 한국어 텍스트
35
+ Preprocessing
36
+
37
+ 텍스트 정제 (최소 길이 필터링)
38
+ BPE 토큰화
39
+ 고정 길이 시퀀스로 분할 (512 tokens)
40
+ 패딩 적용
41
+
42
+ Tokenization
43
+
44
+ Tokenizer: BPE (Byte-Pair Encoding)
45
+ Vocabulary Size: 32,000
46
+ Special Tokens: <pad>, <s>, </s>, <unk>, <mask>
47
+
48
+ Usage
49
+ Loading the Dataset
50
+ pythonfrom datasets import load_dataset
51
+ import pickle
52
+
53
+ # HuggingFace에서 다운로드
54
+ dataset = load_dataset("oz1115/korean-pretraining-corpus")
55
+
56
+ # 또는 로컬 pickle 파일 사용
57
+ with open("train.pkl", "rb") as f:
58
+ train_data = pickle.load(f)
59
+ Example
60
+ python# 첫 번째 샘플
61
+ sample = dataset['train'][0]
62
+ print(sample['input_ids'][:10]) # 첫 10개 토큰
63
+ Related Models
64
+ 이 데이터셋으로 학습된 모델:
65
+
66
+ korean-gpt-150m
67
+
68
+ Citation
69
+ bibtex@misc{korean-pretraining-corpus,
70
+ author = {oz1115},
71
+ title = {korean-pretraining-corpus: Korean Pretraining Corpus},
72
+ year = {2025},
73
+ publisher = {HuggingFace},
74
+ url = {https://huggingface.co/datasets/oz1115/korean-pretraining-corpus}
75
+ }
76
+ Contact
77
+
78
+ HuggingFace: @oz1115
metadata.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "num_train": 130127,
3
+ "num_val": 6849,
4
+ "max_seq_len": 512,
5
+ "vocab_size": 32000,
6
+ "created_at": "2025-10-09T08:12:53.495942"
7
+ }
train.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cab7ce2d6d907b0e9fe6584cfe836558fa8923bdb768811b8d286092c2086d02
3
+ size 180018284
val.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d08e8a89484925fe9a5a18b803820de6b29ab0a104d078c10fde72442051f4d7
3
+ size 9391388