haebo/meow-v1-dataset
Viewer • Updated • 8.78k • 4
How to use haebo/meow-clovax-v1 with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-generation", model="haebo/meow-clovax-v1")
messages = [
{"role": "user", "content": "Who are you?"},
]
pipe(messages) # Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("haebo/meow-clovax-v1")
model = AutoModelForCausalLM.from_pretrained("haebo/meow-clovax-v1")
messages = [
{"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
tokenize=True,
return_dict=True,
return_tensors="pt",
).to(model.device)
outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))How to use haebo/meow-clovax-v1 with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "haebo/meow-clovax-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "haebo/meow-clovax-v1",
"messages": [
{
"role": "user",
"content": "What is the capital of France?"
}
]
}'docker model run hf.co/haebo/meow-clovax-v1
How to use haebo/meow-clovax-v1 with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "haebo/meow-clovax-v1" \
--host 0.0.0.0 \
--port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "haebo/meow-clovax-v1",
"messages": [
{
"role": "user",
"content": "What is the capital of France?"
}
]
}'docker run --gpus all \
--shm-size 32g \
-p 30000:30000 \
-v ~/.cache/huggingface:/root/.cache/huggingface \
--env "HF_TOKEN=<secret>" \
--ipc=host \
lmsysorg/sglang:latest \
python3 -m sglang.launch_server \
--model-path "haebo/meow-clovax-v1" \
--host 0.0.0.0 \
--port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "haebo/meow-clovax-v1",
"messages": [
{
"role": "user",
"content": "What is the capital of France?"
}
]
}'How to use haebo/meow-clovax-v1 with Docker Model Runner:
docker model run hf.co/haebo/meow-clovax-v1
meow-clovax-v1은 감정(emotion)과 동물 유형(post_type)에 따라 문장을 자연스럽게 변환하는 한국어 LLM입니다.
naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-1.5B를 기반으로 Supervised Finetuning(SFT) 방식으로 학습되었습니다.| 항목 | 설명 |
|---|---|
| Base Model | HyperCLOVAX-SEED-Text-Instruct-1.5B |
| Fine-tuning Method | Supervised Finetuning (SFT) |
| Model Type | Decoder-only |
| Language | Korean (primary) |
| Parameters | 1.5B |
| Precision | fp16 / fp32 |
| Version | v1 |
| Framework | Transformers |
| license | hyperclovax-seed |
content, emotion, post_type, transformed_content 필드로 구성된 jsonl 데이터셋from transformers import AutoTokenizer, AutoModelForCausalLM
model_id = "haebo/meow-clovax-v1"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)
content = "짜증났겠네 나도 아침마다 짜증남"
emotion = "angry"
post_type = "cat"
instruction = f"다음 문장을 {post_type}의 {emotion}한 말투로 바꿔줘."
prompt = (
f"### Instruction:\n{example['instruction']}\n"
f"### Input:\n{example['input']}\n"
f"### Output:\n{example['output']}"
)
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=400)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
v1 모델에는 아래와 같은 데이터셋이 사용되었습니다.이 데이터들은 별도의 전처리(클랜징/필터링) 없이 원본 그대로 활용되었습니다.
파인튜닝 시 프롬프트 구조에 맞게 변경되었습니다.
데이터 구조
각 샘플은 아래와 같은 필드로 구성되어 있습니다.
content: 원본 문장 (일상 한국어)emotion: 감정 레이블 (예: happy, sad, angry 등)post_type: 동물 유형 (예: cat, dog)transformed_content: 감정 및 동물 말투로 변환된 문장예시
{
"content": "오늘 점심 뭐 먹지.",
"emotion": "normal",
"post_type": dog",
"transformed_content": "오늘 점심 뭐 먹지멍? 🐾 맛있는 냄새가 나는 것 같다멍! 주인님, 저 밥 어딨냐왈! 빨리 밥그릇 채워달라멍! 🦴 ᓚ₍´ ꒳ `₎ა"
}