README.md

---
license: gemma
datasets:
- MarkrAI/KOpen-HQ-Hermes-2.5-60K
language:
- ko
metrics:
- accuracy
base_model:
- google/gemma-2-2b-it
pipeline_tag: text2text-generation
---
# Gemma-2B Quiz Answering Model

This project fine-tunes the Gemma-2B model to provide answers to quiz-related questions. The model is designed to handle complex problems or quizzes and generate clear and accurate responses in Korean.

## Table of Contents
- [Model Overview](#model-overview)
- [How to Use](#how-to-use)
- [Training Details](#training-details)
- [Model Performance](#model-performance)
- [Limitations and Future Work](#limitations-and-future-work)

## Model Overview
The **Gemma-2B Quiz Answering Model** is built on top of the [Gemma-2B](https://huggingface.co/google/gemma-2b) base model. This version has been fine-tuned to better handle complex quiz questions and generate responses in natural Korean, addressing issues with awkward language generation from the base model.

- **Model Name**: `gemma-2b-quiz-ko`
- **Purpose**: Answer complex quiz and problem-solving questions.
- **Language**: Korean (ko)

## How to Use
You can use the model by loading it from Hugging Face Hub. Below is a simple usage example with the `transformers` library:

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained("DORAEMONG/gemma-2b-quiz-ko")
tokenizer = AutoTokenizer.from_pretrained("DORAEMONG/gemma-2b-quiz-ko")

# Input a quiz question
question = "다음 수학 문제의 답은 무엇입니까? 스피너가 A, B, C로 나뉘어 있을 때..."

inputs = tokenizer(question, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)

# Decode the generated text
print(tokenizer.decode(outputs[0], skip_special_tokens=True))