Instructions to use hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2")
model = AutoModelForCausalLM.from_pretrained("hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2", device_map="auto")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2

SGLang

How to use hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2" \
    --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": "hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

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 "hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2" \
        --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": "hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2 with Docker Model Runner:
```
docker model run hf.co/hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2
```

HyperCLOVA X U-CURATE Random Epoch3 Merged v0.2

Powered by HyperCLOVA X

2026 K-DS 해커톤 예선 및 K-AI 리더보드 평가를 위해 만든 실험용 한국어 fine-tuned model입니다. HyperCLOVA X SEED Text Instruct 1.5B의 고정 revision에 U-CURATE Random QLoRA adapter를 병합했습니다. 별도 base model이나 PEFT adapter 없이 AutoModelForCausalLM과 vLLM에서 직접 읽을 수 있는 BF16 standalone full-model artifact입니다.

모델 정보

Base: naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-1.5B
Base revision: 0728a47d632019a8da5f53b663db1c175dc04115
Architecture: LlamaForCausalLM
Format: BF16 sharded safetensors
Fine-tuning: 4-bit NF4 QLoRA, assistant-only masking
Adapter SHA-256: 55280d552281b2c5659d3a047bcf4ee5bf540ac1f1ff8da0d854eaf8771bc954

학습 요약

Unique training samples: 400 (AIHUB_71533, AIHUB_569)
Epochs: 3
Sample presentations: 1,200
Optimizer steps: 150
Batch 1, gradient accumulation 8, seed 42
Learning rate: 1e-4, cosine scheduler
Formatted tokens: 896,493
Supervised tokens: 12,909

내부 검증

동일한 고정 holdout 200건에서 primary metric은 0.523114였습니다. 이 값은 single-seed two-source proxy 결과이며 공식 K-AI 점수가 아닙니다. 1-epoch 기존 Random baseline의 내부 primary 0.433489와 비교하는 epoch-control 실험입니다.

Transformers

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    dtype=torch.bfloat16,
    device_map="auto",
)

vLLM

vllm serve hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2 --dtype bfloat16

한계

400건, 2개 AI Hub source, single seed 기반의 소규모 실험 모델입니다.
본 모델의 내부 점수 향상이 공식 K-AI 점수 향상을 보장하지 않습니다.
일반 한국어 성능, 사실성, 안전성 및 편향을 보장하지 않습니다.
고위험 의사결정에는 사용하지 마십시오.

라이선스

이 derivative model은 저장소에 포함된 HyperCLOVA X SEED Model License Agreement를 따릅니다. 전체 조건은 LICENSE, 변경 사항은 MODIFICATIONS.md, 고지는 NOTICE를 확인하십시오.

Downloads last month: 324

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for hoySky92/HyperCLOVA-X-U-CURATE-Random-Epoch3-Merged-v0.2

Base model

naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-1.5B

Finetuned

(17)

this model