Downloads · 30 days
4
33% of all-time downloads
HyperCLOVAX-SEED-Text-Instruct-0.5B-naver-actions-lora
naver-ellm/HyperCLOVAX-SEED-Text-Instruct-0.5B-naver-actions-lora
HyperCLOVAX-SEED-Text-Instruct-0.5B-naver-actions-lora is a text generation model from naver-ellm. Use it when you need the model to write or continue text. It is set up for peft.
HyperCLOVAX-SEED-Text-Instruct-0.5B 를 naver-actions function-calling 데이터로 LoRA 파인튜닝한 어댑터. tool 명세를 system content 에 텍스트로 주입하고 정답 함수호출을 assistant content 에 JSON 배열로 생성하는 방식 (멀티콜 지원).
All-time downloads
12
Public
Repo size
39.4 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors39.4 MB · 83%
At a glance
- Task
- Text Generation
- Library
- peft
- Access
- Public
- Created
- Jul 11, 2026
- Updated
- Jul 11, 2026
- SHA
- 15b182f0
Base models
Model card
From the Hugging Face model README
HyperCLOVAX-SEED-Text-Instruct-0.5B naver-actions LoRA
HyperCLOVAX-SEED-Text-Instruct-0.5B 를 naver-actions function-calling 데이터로 LoRA 파인튜닝한 어댑터. tool 명세를 system content 에 텍스트로 주입하고 정답 함수호출을 assistant content 에 JSON 배열로 생성하는 방식 (멀티콜 지원).
학습 설정
- 데이터: naver-actions train 4,230건 (3 epochs, batch 8, lr 2e-4 cosine)
- LoRA: r=16, alpha=32, dropout=0.05, target = q/k/v/o/gate/up/down_proj (tie_word_embeddings 모델이라 embed_tokens/lm_head 제외 → GGUF 변환 가능)
- bf16, MPS (Apple Silicon)
평가 (naver-actions held-out testset 앞 100건, greedy)
| model | overall | name_acc | arg_acc | fc_only | neg_only |
|---|---|---|---|---|---|
| base | 0.120 | 0.120 | 0.120 | 0.000 | 0.923 |
| + LoRA | 0.900 | 0.940 | 0.900 | 0.885 | 1.000 |
지표는 functiongemma-270m-it 파이프라인과 동일 기준 (함수명 리스트 + 인자 dict
리스트 완전일치). 평가 스크립트: ellm 저장소
finetune/HyperCLOVAX-SEED-Text-Instruct-0.5B/eval_compare.py.
사용
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base = AutoModelForCausalLM.from_pretrained(
"naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-0.5B")
model = PeftModel.from_pretrained(
base, "naver-ellm/HyperCLOVAX-SEED-Text-Instruct-0.5B-naver-actions-lora")
학습 때와 동일하게 tool 명세를 system 메시지에 텍스트로 넣어야 함.
GGUF 변환: convert_lora_to_gguf.py --base <base모델> --outtype f16 <이 어댑터>
- Task
- Text Generation
- Library
- peft
- Languages
- ko
- Created
- Jul 11, 2026
- Updated
- Jul 11, 2026