Update README.md

README.md

@@ -24,10 +24,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 8463845
-    num_examples: 45769
-  download_size: 4133304
-  dataset_size: 8463845
+    num_bytes: 1100442974
+    num_examples: 5848375
+  download_size: 571816053
+  dataset_size: 1100442974
 - config_name: CSharp
   features:
   - name: blob_id
@@ -52,10 +52,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 10799062
-    num_examples: 53304
-  download_size: 5229026
-  dataset_size: 10799062
+    num_bytes: 2392066248
+    num_examples: 11425016
+  download_size: 1232015539
+  dataset_size: 2392066248
 - config_name: Cpp
   features:
   - name: blob_id
@@ -80,10 +80,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 25495370
-    num_examples: 131729
-  download_size: 12327512
-  dataset_size: 25495370
+    num_bytes: 3167426435
+    num_examples: 16246746
+  download_size: 1632803797
+  dataset_size: 3167426435
 - config_name: Go
   features:
   - name: blob_id
@@ -112,10 +112,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 1289738
-    num_examples: 5978
-  download_size: 504218
-  dataset_size: 1289738
+    num_bytes: 433053889
+    num_examples: 1917163
+  download_size: 179388495
+  dataset_size: 433053889
 - config_name: Java
   features:
   - name: blob_id
@@ -140,10 +140,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 131964268
-    num_examples: 606520
-  download_size: 65086760
-  dataset_size: 131964268
+    num_bytes: 10292427437
+    num_examples: 44990158
+  download_size: 5291667797
+  dataset_size: 10292427437
 - config_name: JavaScript
   features:
   - name: blob_id
@@ -168,10 +168,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 28682940
-    num_examples: 143959
-  download_size: 13628578
-  dataset_size: 28682940
+    num_bytes: 2654326008
+    num_examples: 13253431
+  download_size: 1287066511
+  dataset_size: 2654326008
 - config_name: Markdown
   features:
   - name: blob_id
@@ -196,10 +196,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 114451402
-    num_examples: 547852
-  download_size: 54229584
-  dataset_size: 114451402
+    num_bytes: 4268378053
+    num_examples: 20687077
+  download_size: 2058772192
+  dataset_size: 4268378053
 - config_name: PHP
   features:
   - name: blob_id
@@ -224,10 +224,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 4305039
-    num_examples: 22365
-  download_size: 1987572
-  dataset_size: 4305039
+    num_bytes: 1985843762
+    num_examples: 9914497
+  download_size: 983498806
+  dataset_size: 1985843762
 - config_name: Python
   features:
   - name: blob_id
@@ -252,10 +252,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 89313081
-    num_examples: 458826
-  download_size: 43846401
-  dataset_size: 89313081
+    num_bytes: 4947575770
+    num_examples: 25286019
+  download_size: 2500795086
+  dataset_size: 4947575770
 - config_name: Ruby
   features:
   - name: blob_id
@@ -280,10 +280,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 722353
-    num_examples: 3862
-  download_size: 308486
-  dataset_size: 722353
+    num_bytes: 592832039
+    num_examples: 2976874
+  download_size: 284535771
+  dataset_size: 592832039
 - config_name: Rust
   features:
   - name: blob_id
@@ -308,10 +308,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 288366
-    num_examples: 1494
-  download_size: 117297
-  dataset_size: 288366
+    num_bytes: 227434676
+    num_examples: 1135379
+  download_size: 103158397
+  dataset_size: 227434676
 - config_name: SQL
   features:
   - name: blob_id
@@ -336,10 +336,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 6647337
-    num_examples: 34493
-  download_size: 3227395
-  dataset_size: 6647337
+    num_bytes: 505669712
+    num_examples: 2504412
+  download_size: 261176608
+  dataset_size: 505669712
 - config_name: Shell
   features:
   - name: blob_id
@@ -364,10 +364,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 1019999
-    num_examples: 5371
-  download_size: 456837
-  dataset_size: 1019999
+    num_bytes: 811611733
+    num_examples: 4133547
+  download_size: 394872047
+  dataset_size: 811611733
 - config_name: Swift
   features:
   - name: blob_id
@@ -392,10 +392,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 8229529
-    num_examples: 37726
-  download_size: 3715589
-  dataset_size: 8229529
+    num_bytes: 529873695
+    num_examples: 2454309
+  download_size: 257883733
+  dataset_size: 529873695
 - config_name: TypeScript
   features:
   - name: blob_id
@@ -420,10 +420,10 @@ dataset_info:
     dtype: large_string
   splits:
   - name: train
-    num_bytes: 4690102
-    num_examples: 22948
-  download_size: 2080678
-  dataset_size: 4690102
+    num_bytes: 904736029
+    num_examples: 4290356
+  download_size: 425942502
+  dataset_size: 904736029
 configs:
 - config_name: C
   data_files:
@@ -486,3 +486,131 @@ configs:
   - split: train
     path: TypeScript/train-*
 ---
+
+[HuggingFaceTB/stack-edu](https://huggingface.co/datasets/HuggingFaceTB/stack-edu) 에서 한국어가 포함된 데이터만 발췌
+
+- 주석이나 소스코드 내부에 한국어가 포함된 경우
+- 한국어로 작성된 문서(Markdown 등)
+
+**특징**
+
+- **한글이 5글자 이상** 들어있으면 한국어가 포함된 데이터라고 판단함
+- 극소수지만 한국어가 아닌데 유니코드가 깨져서 우연히 한글이 검출된 경우도 있는 것 같음
+
+  - [SmolLM2](HuggingFaceTB/SmolLM2-1.7B-Instruct) 토크나이저 기준 토큰 수 비교 (총 125B -> 4.1B, 마크다운 제외 시 2.1B)
+
+| Language   | Stack-Edu (B tokens) | Stack-Edu-Ko (M tokens) |
+|------------|----------------------|-------------------------|
+| Python     | 21.8                 | 430.8                   |
+| Cpp        | 16.0                 | 187.4                   |
+| Markdown   | 14.0                 | 1990.9                  |
+| C          | 11.1                 | 76.5                    |
+| JavaScript | 11.1                 | 179.6                   |
+| Java       | 42.1                 | 810.5                   |
+| SQL        | 9.62                 | 240.1                   |
+| PHP        | 9.07                 | 32.2                    |
+| C-Sharp    | 8.87                 | 66.8                    |
+| TypeScript | 3.03                 | 23.4                    |
+| Shell      | 3.13                 | 8.5                     |
+| Swift      | 1.83                 | 41.6                    |
+| Go         | 1.80                 | 6.5                     |
+| Rust       | 1.75                 | 5.2                     |
+| Ruby       | 1.61                 | 4.2                     |
+
+
+
+
+# 💻 Stack-Edu
+
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/3W_vdVnYBBAifrF5JU5QM.png)
+
+Stack-Edu is a 125B token dataset of educational code filtered from [The Stack v2](https://huggingface.co/datasets/bigcode/the-stack-v2), precisely the curated training corpus of [StarCoder2](https://arxiv.org/abs/2402.19173) models denoted StarCoder2Data. It is intended for Language Models training.
+
+This dataset was curated using a classifier-based filtering strategy, inspired by [📚 FineWeb-Edu](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu), to retain only the highest-quality educational programming content.
+
+Stack-Edu shows consistent improvement over StarCoder2data on all the programming languages on MultiPL-E benchmark.
+
+<img src="https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/GWnPgD0diMu0I8buK6mvG.png" width="600"/>
+
+## Downloading the data
+This dataset only contains the SWHIDs to download the code files and not the content of the files itself. The contents can be downloaded from Software Heritage's S3 bucket to ensure data compliance. 
+Please refer to [the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2-train-full-ids) for the data license.
+
+When running on a 16-core AWS `us-east-1` instance, this script takes ~6 hours to download the files: 
+```python
+import boto3
+import gzip
+from datasets import load_dataset
+from botocore.exceptions import ClientError
+
+num_proc = 16
+s3 = boto3.client('s3')
+bucket_name = "softwareheritage"
+
+def download_contents(blob_id):
+    key = f"content/{blob_id}"
+    try:
+        obj = s3.get_object(Bucket=bucket_name, Key=key)
+        with gzip.GzipFile(fileobj=obj['Body']) as fin:
+            content = fin.read().decode("utf-8", errors="ignore")
+        return {"text": content, "download_success": True}
+    except ClientError as e:
+        if e.response['Error']['Code'] == 'NoSuchKey':
+            print(f"File not found: {key}")
+            return {"text": "", "download_success": False}
+        else:
+            raise
+
+# For Python
+ds = load_dataset("HuggingFaceTB/stack-edu", "Python", split="train", num_proc=num_proc)
+ds = ds.map(download_contents, input_columns="blob_id", num_proc=num_proc)
+
+# Filter out failed downloads
+ds = ds.filter(lambda x: x['download_success'])
+
+# Optionally, print the first example to verify the data
+print(ds[0])
+```
+
+## Details
+The table below shows the number of tokens in each programming language using [SmolLM2](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct) tokenizer.
+
+| Language   | Stack-Edu (B tokens) |
+|------------|----------------------|
+| Python     | 21.8                 |
+| Cpp        | 16.0                 |
+| Markdown   | 14.0                 |
+| C          | 11.1                 |
+| JavaScript | 11.1                 |
+| Java       | 42.1                 |
+| SQL        | 9.62                 |
+| PHP        | 9.07                 |
+| C-Sharp    | 8.87                 |
+| TypeScript | 3.03                 |
+| Shell      | 3.13                 |
+| Swift      | 1.83                 |
+| Go         | 1.80                 |
+| Rust       | 1.75                 |
+| Ruby       | 1.61                 |
+
+
+## Dataset curation
+To build Stack-Edu, we:
+
+- Selected 15 largest programming languages from StarCoder2Data 
+- Trained 15 language-specific classifiers, using [StarEncoder](https://huggingface.co/bigcode/starencoder) model on synthetic annotations generated by Llama3-70B-Instruct. The classifiers for each language are available in this [collection](https://huggingface.co/collections/HuggingFaceTB/the-ultimate-collection-of-code-classifiers-67b5aa3eb8994a4b71453005).
+- Applied a filtering threshold of 3 (out of 5) to retain highly educational content, except for Java, which performed best at threshold 2.
+
+## Citation Information
+
+```
+@misc{allal2025smollm2smolgoesbig,
+      title={SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model}, 
+      author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Gabriel Martín Blázquez and Guilherme Penedo and Lewis Tunstall and Andrés Marafioti and Hynek Kydlíček and Agustín Piqueres Lajarín and Vaibhav Srivastav and Joshua Lochner and Caleb Fahlgren and Xuan-Son Nguyen and Clémentine Fourrier and Ben Burtenshaw and Hugo Larcher and Haojun Zhao and Cyril Zakka and Mathieu Morlon and Colin Raffel and Leandro von Werra and Thomas Wolf},
+      year={2025},
+      eprint={2502.02737},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2502.02737}, 
+}
+```