solar-kor-resume

Update @ 2024.05.27: First release of Ocelot-Ko-self-instruction-10.8B-v1.0

This model card corresponds to the 10.8B Instruct version of the Solar-Ko model.

The train wad done on A100-80GB

Resources and Technical Documentation:

Citation

@misc {cpm-ai/Ocelot-Ko-self-instruction-10.8B-v1.0,
    author       = { {frcp, nebchi} },
    title        = { solar-kor-resume},
    year         = 2024,
    url          = { https://huggingface.co/cpm-ai/Ocelot-Ko-self-instruction-10.8B-v1.0 },
    publisher    = { Hugging Face }
}

Model Developers: frcp, nebchi

Model Information

Resume Proofreading and evaluation of inputs and outputs.

Description

It has been trained with a large amount of Korean tokens compared to other LLMs, enabling it to generate high-quality Korean text.

Model Architecture Solar is an auto-regressive language model that is scaled using the DUS method.

*You can find dataset list here: https://huggingface.co/datasets/cpm-ai/gpt-self-introduction-all

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.

Running the model on a single / multi GPU

# pip install accelerate, flash_attn, sentencepiece
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("cpm-ai/Ocelot-Ko-self-instruction-10.8B-v1.0")
model = AutoModelForCausalLM.from_pretrained("cpm-ai/Ocelot-Ko-self-instruction-10.8B-v1.0", device_map="auto")

pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=4096, streamer=streamer)

text = λ„ˆλŠ” μžκΈ°μ†Œκ°œμ„œ 첨삭 μ „λ¬Έκ°€μ•Ό.
μ£Όμ–΄μ§„ μžκΈ°μ†Œκ°œμ„œλ₯Ό μ²¨μ‚­ν•΄μ„œ λ‹€μ‹œ μž‘μ„±ν•΄μ•Όν•΄.
좜λ ₯ν˜•μ‹μ€ λ‹€μŒμ„ μ§€μΌœμ•Όν•΄.

[첨삭]

λ‹€μŒμ΄ μžκΈ°μ†Œκ°œμ„œμ•Ό :
[μ €λŠ” μ–΄λ¦° μ‹œμ ˆλΆ€ν„° μ™„λ²½μ£Όμ˜μ μΈ 성격을 κ°€μ§€κ³  μžˆμ—ˆμŠ΅λ‹ˆλ‹€. 이둜 인해 항상 μžμ‹ μ˜ λŠ₯λ ₯에 λŒ€ν•œ λΆˆμ•ˆκ°μ„ 느끼며 κ³Όλ„ν•œ 슀트레슀λ₯Ό λ°›μ•„μ™”μŠ΅λ‹ˆλ‹€. ν•™μ°½ μ‹œμ ˆμ—λŠ” κ³Όμ œλ‚˜ ν”„λ‘œμ νŠΈλ₯Ό μ™„λ²½ν•˜κ²Œ λ§ˆλ¬΄λ¦¬ν•˜μ§€ λͺ»ν•˜λ©΄ 자쑴감이 크게 ν”λ“€λ ΈμŠ΅λ‹ˆλ‹€. 쀑학ꡐ μ‹œμ ˆμ—λŠ” ν•œ κ°€μ§€ λ¬Έμ œμ— λ„ˆλ¬΄ 였랜 μ‹œκ°„μ„ νˆ¬μžν•˜μ—¬ λ‹€λ₯Έ ν•™μŠ΅ 기회λ₯Ό λ†“μΉ˜κΈ°λ„ ν–ˆμŠ΅λ‹ˆλ‹€. μ΄λŸ¬ν•œ κ²½ν—˜λ“€μ€ μ €μ—κ²Œ 완벽함을 μΆ”κ΅¬ν•˜λŠ” 것이 μ’…μ’… ν˜„μ‹€μ— λΆ€μ ν•©ν•˜λ‹€λŠ” 것을 κΉ¨λ‹¬κ²Œ ν–ˆμŠ΅λ‹ˆλ‹€.

고등학ꡐ와 λŒ€ν•™κ΅μ— μ§„ν•™ν•˜λ©΄μ„œλ„ μ΄λŸ¬ν•œ μ™„λ²½μ£Όμ˜μ μΈ 성격을 κ·Ήλ³΅ν•˜κΈ° μœ„ν•΄ λ…Έλ ₯ν–ˆμŠ΅λ‹ˆλ‹€. ν•˜μ§€λ§Œ μ—¬μ „νžˆ μ‹€νŒ¨λ₯Ό λ°›μ•„λ“€μ΄λŠ” 것이 μ–΄λ ΅κ³ , μžμ‹ μ˜ ν•œκ³„λ₯Ό μΈμ •ν•˜λŠ” 것이 μ–΄λ €μ› μŠ΅λ‹ˆλ‹€. μ΄λŸ¬ν•œ 과정을 톡해 완벽함에 λŒ€ν•œ 강박이 μ €μ˜ μ„±μž₯과정을 μ œμ•½ν•˜λŠ” μš”μΈμ΄ λ˜μ—ˆμŒμ„ κΉ¨λ‹¬μ•˜μŠ΅λ‹ˆλ‹€.]"""

messages = [
    {
        "role": "user",
        "content": "{}".format(text)
    }
]

prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

outputs = pipe(
    prompt,
    temperature=0.2,
    add_special_tokens=True
)
print(outputs[0]["generated_text"][len(prompt):])

results

[첨삭]

μ–΄λ¦° μ‹œμ ˆλΆ€ν„° μ €λŠ” μ™„λ²½ν•œ κ²°κ³Όλ₯Ό μΆ”κ΅¬ν•˜λ©° 슀슀둜λ₯Ό μ••λ°•ν•΄μ˜¨ μ„±κ²©μ΄μ—ˆμŠ΅λ‹ˆλ‹€. μ΄λŠ” ν•™μ—…κ³Ό κ΄€λ ¨λœ 슀트레슀둜 이어져, 과제λ₯Ό μ™„μˆ˜ν•˜λ”λΌλ„ λ§Œμ‘±λ³΄λ‹€λŠ” 뢈만쑱의 감정이 더 컸던 μ‹œκΈ°μ˜€μŠ΅λ‹ˆλ‹€. 특히 쀑학ꡐ λ•Œ ν•œ 문제λ₯Ό ν•΄κ²°ν•˜κΈ° μœ„ν•΄ μ§€λ‚˜μΉ˜κ²Œ μ˜€λž«λ™μ•ˆ 맀달렀 κ²°κ΅­ μ€‘μš”ν•œ μ‹œκΈ°λ₯Ό λ†“μΉœ κ²½ν—˜μ€ 제 μ„±μž₯에 큰 영ν–₯을 λ―Έμ³€μŠ΅λ‹ˆλ‹€. 이 κ³Όμ •μ—μ„œ μ™„λ²½μ£Όμ˜λ₯Ό μΆ”κ΅¬ν•˜λŠ” 것이 ν˜„μ‹€μ μ΄μ§€ μ•Šμ„ 수 μžˆλ‹€λŠ” 사싀을 κΉ¨λ‹«κΈ° μ‹œμž‘ν–ˆμŠ΅λ‹ˆλ‹€.

고등학ꡐ와 λŒ€ν•™μ—μ„œλŠ” μ΄λŸ¬ν•œ μ„±ν–₯을 κ°œμ„ ν•˜κ³ μž λ‹€μ–‘ν•œ λ…Έλ ₯을 κΈ°μšΈμ˜€μŠ΅λ‹ˆλ‹€. 예λ₯Ό λ“€μ–΄, λͺ©ν‘œλ₯Ό μ„ΈλΆ„ν™”ν•˜κ³  λ‹¨κ³„λ³„λ‘œ μ ‘κ·Όν•˜λ©΄μ„œ 성취감과 μžμ‹ κ°μ„ ν‚€μš°κΈ° μœ„ν•΄ λ…Έλ ₯ν–ˆμŠ΅λ‹ˆλ‹€. λ˜ν•œ, νŒ€ ν”„λ‘œμ νŠΈμ—μ„œ 역할을 λΆ„λ‹΄ν•˜κ³  ν˜‘λ ₯ν•¨μœΌλ‘œμ¨ 개인의 ν•œκ³„λ³΄λ‹€ 전체 μ„±κ³Όλ₯Ό μš°μ„ μ‹œν•˜λŠ” 법을 λ°°μ› μŠ΅λ‹ˆλ‹€. 비둝 아직 μ™„λ²½ν•¨μ΄λΌλŠ” κ΅΄λ ˆλ‘œλΆ€ν„° μ™„μ „νžˆ μžμœ λ‘œμ›Œμ§€μ§€λŠ” λͺ»ν–ˆμ§€λ§Œ, 이λ₯Ό κ·Ήλ³΅ν•˜κ³  μ„±μž₯ν•  수 μžˆλŠ” 방법을 μ°Ύμ•˜λ‹€λŠ” μ μ—μ„œ μžλΆ€μ‹¬μ„ λŠλ‚λ‹ˆλ‹€.

Evaluation Results - LogicKor

Evaluation Results-LogicKor
Model κΈ€μ“°κΈ° 이해 문법
HyperClovaX 8.50 9.50 8.50
solar-1-mini-chat 8.50 7.00 5.21
allganize/Llama-3-Alpha-Ko-8B-Instruct 8.50 8.35 4.92
Synatra-kiqu-7B 4.42 5.71 4.50
Ocelot-ko-10.8B 8.57 7.00 6.57

Evaluation Results - Kobest

λͺ¨λΈ 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
Ocelot-ko-10.8B 72.5    75.9 50.0    51.4 75.8    82.5 91.7    93.8

Software

Training was done using QLoRA

Downloads last month
4
Safetensors
Model size
11B params
Tensor type
BF16
Β·
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for cpm-ai/Ocelot-Ko-self-instruction-10.8B-v1.0

Quantizations
3 models