README.md

---
language:
- ko
datasets:
- kyujinpy/OpenOrca-KO
library_name: transformers
pipeline_tag: text-generation
license: cc-by-nc-4.0
---

# **Korean-OpenOrca-13B**  
![img](./Korean-OpenOrca.png)  

## Model Details

**Model Developers** Kyujin Han (kyujinpy)

**Input** Models input text only.

**Output** Models generate text only.

**Model Architecture**  
Korean-OpenOrca-13B is an auto-regressive language model based on the LLaMA2 transformer architecture.

**Repo Link**  
Github Korean-OpenOrca: (Coming soon...) 

**Base Model**  [hyunseoki/ko-en-llama2-13b](https://huggingface.co/hyunseoki/ko-en-llama2-13b)   

**Training Dataset**  
I use [OpenOrca-KO](https://huggingface.co/datasets/kyujinpy/OpenOrca-KO).  
Using DeepL, translate about [OpenOrca](https://huggingface.co/datasets/Open-Orca/OpenOrca).

I use A100 GPU 40GB and COLAB, when trianing.


# **Model Benchmark**

## KO-LLM leaderboard
- Follow up as [Open KO-LLM LeaderBoard](https://huggingface.co/spaces/upstage/open-ko-llm-leaderboard).  
  
![img](./leaderboard.png)
| Model | Average |Ko-ARC | Ko-HellaSwag | Ko-MMLU | Ko-TruthfulQA | Ko-CommonGen V2 |
| --- | --- | --- | --- | --- | --- | --- |
| Korean-OpenOrca-13B(ours) | NaN | NaN | NaN | NaN | NaN | NaN |  
| [KoT-Platypus2-13B](https://huggingface.co/kyujinpy/KoT-platypus2-13B) | 49.55 | 43.69 | 53.05 | 42.29 | 43.34 | 65.38 |  
| [KO-Platypus2-13B](https://huggingface.co/kyujinpy/KO-Platypus2-13B) | 47.90 | 44.20 | 54.31 | 42.47 | 44.41 | 54.11 | 
| [hyunseoki/ko-en-llama2-13b](https://huggingface.co/hyunseoki/ko-en-llama2-13b) | 46.68 | 42.15 | 54.23 | 38.90 | 40.74 | 57.39 |
| [MarkrAI/kyujin-CoTy-platypus-ko-12.8b](https://huggingface.co/MarkrAI/kyujin-CoTy-platypus-ko-12.8b) | 46.44 | 34.98 | 49.11 | 25.68 | 37.59 | 84.86 | 
  
> Compare with Top 4 SOTA models. (update: 10/09)

  
# Implementation Code
```python
### KO-Platypus
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "kyujinpy/Korean-OpenOrca-13B"
OpenOrca = AutoModelForCausalLM.from_pretrained(
        repo,
        return_dict=True,
        torch_dtype=torch.float16,
        device_map='auto'
)
OpenOrca_tokenizer = AutoTokenizer.from_pretrained(repo)
```

---