Dataset · kyujinpy
Korean OpenOrca Sample Translation
kyujinpy/OpenOrca-KO
About 20,000 instruction examples and chain-of-thought (CoT) reasoning translated into Korean via DeepL.
What's inside
The dataset is a sample of approximately 20,000 text records from the well-known OpenOrca corpus translated into Korean using the DeepL Pro API. It compiles data from FLAN (9,434), T0 (6,351), KoCoT (2,159), CoT (2,117), and NIV (1,571) subsets. Suitable for a wide range of NLP tasks from summarization to question answering.
What you can do with it
- Instruction fine-tuning (SFT) of language models in Korean
- Training step-by-step reasoning models (Chain-of-Thought) for the Korean language
- Tasks
- Text classification, Token classification, Table question answering, Question answering, Zero shot classification, Summarization
- Languages
- Korean
- Size
- 10K–100K rows
- Downloads this month
- 88
- Likes
- 32
Models trained on it
- TheBloke/PiVoT-0.1-Evil-a-GGUFText generation 957
- TheBloke/PiVoT-0.1-early-GGUFText generation 647
- mradermacher/KoR-Orca-Platypus-13B-i1-GGUF 613
- mradermacher/DataVortexTL-1.1B-v0.1-i1-GGUFText generation 610
- mradermacher/Llama-3-Ko-8B-OpenOrca-i1-GGUF 515
- tensorblock/Llama-3-Ko-8B-Instruct-AOG-GGUFText generation 356
- mradermacher/DataVortexTL-1.1B-v0.1-GGUFText generation 265
- mradermacher/Llama-3-Ko-8B-OpenOrca-GGUF 187
How to load it
from datasets import load_dataset
ds = load_dataset("kyujinpy/OpenOrca-KO")