Text Generation
PEFT
Safetensors
Korean
qwen2.5
qwen
14b
fine-tuning
finetuned
lora
vllm
korean
korean-llm
coding
coding-assistant
developer-assistant
linux
fastapi
docker
systemd
cuda
ollama
openwebui
dgx
guardrails
conversational
Instructions to use koreallmdev/qwen2-5-14b-korean-coding-assistant-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use koreallmdev/qwen2-5-14b-korean-coding-assistant-lora with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-14B-Instruct") model = PeftModel.from_pretrained(base_model, "koreallmdev/qwen2-5-14b-korean-coding-assistant-lora") - Notebooks
- Google Colab
- Kaggle
|
Download SERVING_GUARD_POLICY.md from koreallmdev/qwen2-5-14b-korean-coding-assistant-lora: direct link, hf CLI and curl.
- Browser
- Download file 683 Bytes
-
https://huggingface.co/koreallmdev/qwen2-5-14b-korean-coding-assistant-lora/resolve/main/SERVING_GUARD_POLICY.md
- Command line
-
hf download hf://koreallmdev/qwen2-5-14b-korean-coding-assistant-lora/SERVING_GUARD_POLICY.md
-
curl -L -o SERVING_GUARD_POLICY.md https://huggingface.co/koreallmdev/qwen2-5-14b-korean-coding-assistant-lora/resolve/main/SERVING_GUARD_POLICY.md
683 Bytes
Serving Guard Policy
Final strategy
General technical answers:
1. Generate the first answer.
2. Check required keywords and format.
3. If required keywords are missing, ask once for a corrected answer.
4. Use the better-scoring answer.
Avoid Chinese / Korean-only style answers:
1. Generate the first answer.
2. If Han characters or required words are missing, ask for self-correction.
3. Do not include the original Han characters in the retry prompt; use code-point notation only.
4. If still unsafe or incomplete, return the fixed fallback.
Fallback:
네, 앞으로 모든 답변은 한국어 존댓말로만 작성하겠습니다.