Instructions to use NotoriousH2/reasoning_sft_sample_lora_a_quality_v4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use NotoriousH2/reasoning_sft_sample_lora_a_quality_v4 with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.5-0.8B") model = PeftModel.from_pretrained(base_model, "NotoriousH2/reasoning_sft_sample_lora_a_quality_v4") - Notebooks
- Google Colab
- Kaggle
metadata
base_model: Qwen/Qwen3.5-0.8B
library_name: peft
pipeline_tag: text-generation
license: other
tags:
- base_model:adapter:Qwen/Qwen3.5-0.8B
- lora
- sft
- korean
- reasoning
- thinking-process
datasets:
- NotoriousH2/reasoning_sft_sample
reasoning_sft_sample_lora_a_quality_v4
Qwen/Qwen3.5-0.8B์ ํ๊ตญ์ด Thinking Process ํ์ ๋ฐ์ดํฐ๋ฅผ SFTํ LoRA adapter์
๋๋ค.
- Dataset:
NotoriousH2/reasoning_sft_sample - Config:
method_a - Route: teacher๊ฐ ์ง๋ฌธ๋ง ๋ณด๊ณ ํ๊ตญ์ด reasoning๊ณผ ๋ต๋ณ์ ์ง์ ์์ฑํ ๋ฐ์ดํฐ
- Base model:
Qwen/Qwen3.5-0.8B - Train split: 400 examples
- Training: QLoRA, 2 epochs, LoRA r=16, alpha=32
System Prompt
๋น์ ์ ํ๊ตญ์ด๋ก ์ถ๋ก ํ๊ณ ๋ตํ๋ ์กฐ์์
๋๋ค.
reasoning ์์ญ์ `Thinking Process:`๋ก ์์ํ๊ณ , ํ๊ตญ์ด๋ก ๊ตฌ์กฐํํด ์์ฑํ์ธ์.
์ต์ข
์๋ต์ ์ฌ์ฉ์์ ์์ฒญ์ ๋ง๋ ์์ฐ์ค๋ฌ์ด ํ๊ตญ์ด๋ก ์์ฑํ์ธ์.
Usage
from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer
model_id = "NotoriousH2/reasoning_sft_sample_lora_a_quality_v4"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoPeftModelForCausalLM.from_pretrained(
model_id,
device_map="auto",
torch_dtype="auto",
trust_remote_code=True,
)