Kor-Resume-Orion-14B
Update @ 2024.03.28: First release of Kor-Resume-Orion-14B
This model card corresponds to the 14B base version of the Orion-Ko model.
Resources and Technical Documentation:
Citation
@misc {kor-resume-Orion-14B,
author = { {nebchi} },
title = { kor-resume-Orion-14B },
year = 2024,
url = { https://huggingface.co/nebchi/kor-resume-Orion-14B },
publisher = { Hugging Face }
}
Model Developers: nebchi
Model Information
Resume Proofreading and evaluation of inputs and outputs.
Description
The Orion model is trained on 2.5T tokens and supports languages including Korean, Japanese, Chinese, and English. It has been trained with a large amount of Korean tokens compared to other LLMs, enabling it to generate high-quality Korean text. Additionally, it shows improved performance with less data compared to other LLM models.
Running the model on a single / multi GPU
# pip install accelerate, flash_attn, sentencepiece
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("nebchi/kor-resume-Orion-14B")
model = AutoModelForCausalLM.from_pretrained("nebchi/kor-resume-Orion-14B", device_map="auto")
model.generation_config = GenerationConfig.from_pretrained(
"nebchi/kor-resume-Orion-14B"
)
messages = [
{"role": "user", "content": """μ§μλκΈ°λ μ λ λ°μ΄λ λΆμλ ₯κ³Ό λ¬Έμ ν΄κ²° λ₯λ ₯μ μ§λκ³ μμ΅λλ€. 볡μ‘ν μν©μμλ λ
Όλ¦¬μ μΌλ‘ μ κ·Όνμ¬ μ΅μ μ ν΄κ²°μ±
μ μ°Ύμλ΄λ©°, λ°μ΄ν°μ κΉμ ν΅μ°°λ ₯μ λ°νν©λλ€. μ΄λ¬ν μλμ KB κ΅λ―ΌμΉ΄λμ λ°μ΄ν° λΆμ μ
무μ ν° κ°μΉλ₯Ό μ 곡ν κ²μ
λλ€.
νμ§λ§ λλ‘λ μλ²½ν¨μ μΆκ΅¬νλ μ±κ²© νμ μμ
μκ°μ΄ λμ΄λ μ μμ΅λλ€. μ΄ λλ¬Έμ μ λ΅μ μΈ μ
무 κ³νμ΄ νμν μν©μμ μ€μν λΆλΆμ μΆ©λΆν μκ°μ ν μ νμ§ λͺ»ν μ μμ΅λλ€. μ΄λ₯Ό 극볡νκΈ° μν΄ μ μμ μκ² μ μ°μ±μ λΆμ¬νκ³ μμ
μ°μ μμλ₯Ό λͺ
ννκ² μ€μ νλ λ°©λ²μ μ΅νκ³ μμ΅λλ€.
ν λ² νλ‘μ νΈ μ€ μ΄λ €μ΄ λ°μ΄ν° ν¨ν΄μ λΆμν΄μΌ νμ λ, μ λ°μ΄λ λΆμλ ₯μ λ°ννμ¬ λ°μ΄ν° κ°μ μ°κ΄μ±μ μ°Ύμλμ΅λλ€. μ΄ λλ¬Έμ κΈ°μ‘΄ λ°©μμμ λ²μ΄λ μλ‘μ΄ μΈμ¬μ΄νΈλ₯Ό μ»μ μ μμκ³ , νλ‘μ νΈ κ²°κ³Όμ κΈμ μ μΈ μν₯μ λ―Έμ³€μ΅λλ€. κ·Έλ¬λ μ΄μ λμμ νλ‘μ νΈ μΌμ μ΄ λ¦μ΄μ§λ μν©λ μμλλ°, μ΄λ₯Ό 극볡νκΈ° μν΄ λ ν¨μ¨μ μΈ μμ
κ³νμ μΈμ°λ € λ
Έλ ₯νμ¬ κ²°κ³Όμ μΌλ‘ νλ‘μ νΈλ₯Ό μ±κ³΅μ μΌλ‘ λ§λ¬΄λ¦¬ν μ μμμ΅λλ€. μ΄λ κ² μΌλλ°, λΆμ‘±ν λΆλΆ 첨μν΄μ£Όκ³ , 첨μν μ΄μ μ€λͺ
ν΄μ€ μ μμ΄? """}
]
response = finetune_model.chat(tokenizer, messages, streaming=False)
print(response)
results
μ§μλκΈ°μμλ μμ μ κ°μ μΈ λ°μ΄λ λΆμλ ₯κ³Ό λ¬Έμ ν΄κ²° λ₯λ ₯μ κ°μ‘°νλ©°, μ΄λ₯Ό λ°νμΌλ‘ KB κ΅λ―ΌμΉ΄λμ λ°μ΄ν° λΆμ μ
무μ μ ν©νλ€κ³ μ£Όμ₯ν©λλ€. νμ§λ§ λ¨μ μΌλ‘λ μλ²½ν¨μ μΆκ΅¬νλ μ±κ²©μΌλ‘ μΈν΄ μμ
μκ°μ΄ λμ΄λ μ μλ€κ³ μΈκΈνλ©°, μ΄λ₯Ό 극볡νκΈ° μν΄ μ μ°μ±κ³Ό μμ
μ°μ μμ μ€μ μ μ΅νκ³ μλ€κ³ λ§ν©λλ€. λ€μ μμμμλ μ΄λ €μ΄ λ°μ΄ν° ν¨ν΄μ λΆμν κ²½νμ 곡μ νλ©°, μλ‘μ΄ μΈμ¬μ΄νΈλ₯Ό μ»μ΄λ΄λ λ± κΈμ μ μΈ κ²°κ³Όλ₯Ό κ°μ Έμμ§λ§, νλ‘μ νΈ μΌμ μ΄ λ¦μ΄μ§λ μ΄λ €μλ μμλ€κ³ ν©λλ€. μ΄λ₯Ό 극볡νκΈ° μν΄ λ ν¨μ¨μ μΈ μμ
κ³νμ μΈμ μ±κ³΅μ μΌλ‘ νλ‘μ νΈλ₯Ό λ§λ¬΄λ¦¬νλ€κ³ ν©λλ€.
첨μν μ΄μ λ λ€μκ³Ό κ°μ΅λλ€. λ¨Όμ , μ§μλκΈ°μμλ μμ μ κ°μ μ κ°μ‘°νλ©΄μ KB κ΅λ―ΌμΉ΄λμ λ°μ΄ν° λΆμ μ
무μ μ ν©νλ€λ κ²μ μ£Όμ₯νκ³ μμ΅λλ€. νμ§λ§ μ΄ λΆλΆμ΄ ꡬ체μ μΌλ‘ μ΄λ€ μλμ λ§νλ κ²μΈμ§ λͺ
ννμ§ μμ΅λλ€. λ°λΌμ "λ°μ΄λ λΆμλ ₯κ³Ό λ¬Έμ ν΄κ²° λ₯λ ₯"μ΄λΌλ ννμ μ’ λ ꡬ체μ μΌλ‘ μ€λͺ
νκ±°λ, ν΄λΉ λΆμΌμμμ κ²½νμ΄λ μ±κ³Όλ₯Ό μ μνλ κ²μ΄ μ’μ΅λλ€.
λν, λ¨μ λΆλΆμμλ μλ²½ν¨μ μΆκ΅¬νλ μ±κ²©μΌλ‘ μΈν΄ μμ
μκ°μ΄ λμ΄λ μ μλ€λ κ²μ μΈκΈνκ³ μμ§λ§, μ΄ λΆλΆμ΄ μ§μμμ μ
무 μνμ μ΄λ€ μν₯μ λ―ΈμΉλμ§μ λν ꡬ체μ μΈ μ€λͺ
μ΄ μμ΅λλ€. λ°λΌμ μ΄ λ¨μ μ΄ μ΄λ»κ² μ
무 μνμ μν₯μ λ―ΈμΉλμ§, κ·Έλ¦¬κ³ μ΄λ₯Ό 극볡νκΈ° μν΄ μ΄λ€ λ
Έλ ₯μ νκ³ μλμ§μ λν΄ λ μμΈν μ€λͺ
νλ κ²μ΄ μ’μ΅λλ€.
λ€μμΌλ‘, μμμμλ μ΄λ €μ΄ λ°μ΄ν° ν¨ν΄μ λΆμν κ²½νμ 곡μ νκ³ μμ΅λλ€. μ΄ κ²½νμ μ§μμμ λΆμλ ₯κ³Ό λ¬Έμ ν΄κ²° λ₯λ ₯μ 보μ¬μ€ μ μλ μ’μ μ¬λ‘μ
λλ€. λ€λ§, μ΄ κ²½νμ΄ μ§μμμ μλκ³Ό μ΄λ€ κ΄λ ¨μ΄ μλμ§, κ·Έλ¦¬κ³ μ΄ κ²½νμ ν΅ν΄ μ΄λ€ κ²μ λ°°μ λμ§λ₯Ό λ μμΈν μ€λͺ
νλ κ²μ΄ μ’μ΅λλ€. λν, νλ‘μ νΈ μΌμ μ΄ λ¦μ΄μ§λ μ΄λ €μμ 극볡νκΈ° μν΄ μ΄λ€ λ°©λ²μ μ¬μ©νλμ§μ λν ꡬ체μ μΈ μ€λͺ
μ΄ μμΌλ©΄ λμ± ν¨κ³Όμ μΌ κ²μ
λλ€.
λ§μ§λ§μΌλ‘, μμμμλ νλ‘μ νΈλ₯Ό μ±κ³΅μ μΌλ‘ λ§λ¬΄λ¦¬νλ€λ κ²μ μΈκΈνκ³ μμ΅λλ€. νμ§λ§ μ΄ λΆλΆμ΄ μ΄λ€ μ±κ³Όλ₯Ό κ°μ ΈμμΌλ©°, μ΄λ₯Ό ν΅ν΄ μ΄λ€ κ²μ λ°°μ λμ§λ₯Ό λ μμΈν μ€λͺ
νλ κ²μ΄ μ’μ΅λλ€.
λ°λΌμ, μ§μλκΈ°μμλ μμ μ κ°μ κ³Ό λ¨μ μ ꡬ체μ μΌλ‘ μ€λͺ
νκ³ , μμλ₯Ό ν΅ν΄ μ΄λ₯Ό λ·λ°μΉ¨νλ κ²μ΄ μ’μ΅λλ€. λν, μμμμλ ꡬ체μ μΈ μ±κ³Όμ κ²½νμ μ μνμ¬ μ§μμμ μλκ³Ό κ²½νμ λμ± κ°μ‘°νλ κ²μ΄ μ’μ΅λλ€.
Inputs and outputs
- Input: Text string, such as a question, a prompt, or a document to be Proofreaded.
- Output: Generated Korea text in response to the input, such as an answer to a question, or a evaluation of a resume.
Software
Training was done using QLoRA
Evaluation Results
| λͺ¨λΈ λͺ μΉ | Average n=0 n=5 |
HellaSwag n=0 n=5 |
COPA n=0 n=5 |
BooIQ n=0 n=5 |
|---|---|---|---|---|
| KoGPT | 58.2 63.7 | 55.9 58.3 | 73.5 72.9 | 45.1 59.8 |
| Polyglot-ko-13B | 62.4 68.2 | 59.5 63.1 | 79.4 81.1 | 48.2 60.4 |
| LLaMA 2-13B | 45.2 60.5 | 41.3 44.0 | 59.3 63.8 | 34.9 73.8 |
| Baichuan 2-13B | 52.7 53.9 | 39.2 39.6 | 60.6 60.6 | 58.4 61.5 |
| QWEN-14B | 47.8 66.4 | 45.3 46.8 | 64.9 68.9 | 33.4 83.5 |
| Orion-14B-Chat | 68.8 73.2 | 47.0 49.6 | 77.7 79.4 | 81.6 90.7 |
| kor-resume-Orion | 69.7 74.6 | 48.2 51.2 | 77.9 81.3 | 83.1 91.2 |
Declarations, License
Declarations
We strongly urge all users not to use the Orion-14B model for any activities that may harm national or social security or violate the law. Additionally, we request users not to use the Orion-14B model for internet services without proper security review and filing. We hope all users abide by this principle to ensure that technological development takes place in a regulated and legal environment. We have done our best to ensure the compliance of the data used in the model training process. However, despite our significant efforts, unforeseen issues may still arise due to the complexity of the model and data. Therefore, if any problems arise due to the use of the Orion-14B open-source model, including but not limited to data security issues, public opinion risks, or any risks and issues arising from the model being misled, abused, disseminated, or improperly utilized, we will not assume any responsibility.
License
Community use of the Orion-14B series models
- For code, please comply with Apache License Version 2.0
- For model, please comply with γOrion-14B Seriesγ Models Community License Agreement
- Downloads last month
- 14