Datasets:
metadata
language:
- tr
- en
license: cc-by-4.0
task_categories:
- text-generation
tags:
- turkish
- continual-pretraining
- CPT
- wikipedia
- fineweb
- c4
size_categories:
- 1B<n<10B
Turkish CPT Dataset
A high-quality Turkish + English dataset for Continued Pre-Training (CPT) of language models.
Dataset Summary
| Property | Value |
|---|---|
| Total examples | 1,908,378 |
| Total tokens | ~2.19B |
| Turkish ratio | ~80% |
| English ratio | ~20% |
| Languages | Turkish, English |
Sources
| Source | Language | Examples | Description |
|---|---|---|---|
| wikimedia/wikipedia (tr) | TR | ~534K | Turkish Wikipedia |
| wikimedia/wikipedia (en) | EN | ~134K | English Wikipedia (20% replay) |
| habanoz/c4_tr_fineweb_plus | TR | ~500K | Filtered Turkish web text |
| HuggingFaceFW/fineweb-2 (tur_Latn) | TR | ~500K | High-quality Turkish web data |
| HuggingFaceFW/fineweb | EN | ~300K | High-quality English web data |
Cleaning Pipeline
Applied industry-standard cleaning steps:
- UTF-8 NFC normalization — unicode noise removal
- Whitespace normalization — excess newlines, tabs, spaces
- URL removal — web boilerplate
- Alphanumeric ratio filter — spam/symbol detection (min 50%)
- Repetitive line filter — boilerplate detection (min 30% unique lines)
- Minimum 50 tokens — very short text removal
- Maximum 100K tokens — abnormally long document removal
60,357 examples removed (3.1%), 1.3% token loss.
English Replay
20% English data is mixed in following best practices from continual pretraining research to prevent catastrophic forgetting of the base model's reasoning capabilities.
Usage
from datasets import load_dataset
dataset = load_dataset("afkfatih/turkish-cpt-dataset", split="train")
Intended Use
This dataset is intended for CPT of instruction-tuned or base language models to improve Turkish language understanding and generation while preserving English capabilities.
License
Dataset components inherit their original licenses. Compiled dataset released under CC-BY-4.0.