Carrot Llama-3.2 Rabbit Ko Open-Source Large Language Model - Supports both Korean and English, excels in text generation.

Llama 3.2 Rabbit Ko 3B Instruct

Developed by CarrotAI

Carrot Llama-3.2 Rabbit Ko is a large language model fine-tuned with instructions, supporting Korean and English, and performs excellently in text generation tasks.

Large Language Model

Safetensors

Supports Multiple Languages#Korean-English bilingual instruction fine-tuning #Lightweight with 3B parameters #Optimized for Korean

Downloads 2,169

Release Time : 9/30/2024

Model Overview

This model is a large language model with a 3B parameter scale based on the Llama-3.2 architecture. It has been optimized through instruction fine-tuning and is particularly good at text generation tasks in Korean and English.

Model Features

Multilingual support

Specifically optimized the text generation ability in Korean and English

Instruction fine-tuning

After supervised fine-tuning (SFT), it can better understand and execute instructions

Lightweight

With a 3B parameter scale, it reduces resource requirements while maintaining performance

Model Capabilities

Korean text generation

English text generation

Instruction understanding and execution

Question-answering system

Use Cases

Education

Korean learning assistant

Help learners generate Korean learning materials and practice questions

Customer service

Multilingual customer service robot

Provide automatic customer service responses in Korean and English

🚀 Carrot Llama-3.2 Rabbit Ko

Carrot Llama-3.2 Rabbit Ko is an instruction-tuned large language model supporting Korean and English, offering high - quality text generation capabilities.

image/webp

🚀 Quick Start

💻 Usage Examples

Basic Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CarrotAI/Llama-3.2-Rabbit-Ko-3B-Instruct")
tokenizer = AutoTokenizer.from_pretrained("CarrotAI/Llama-3.2-Rabbit-Ko-3B-Instruct")

✨ Features

Model Details

Name: Carrot Llama-3.2 Rabbit Ko
Version: 3B Instruct
Base Model: CarrotAI/Llama-3.2-Rabbit-Ko-3B-Instruct
Languages: Korean, English
Model Type: Large Language Model (Instruction-tuned)

Training Process

This model has gone through the following main training steps:

SFT (Supervised Fine-Tuning)
- Fine-tuned the base model using high-quality Korean and English datasets.

Limitations

Limited performance in complex tasks due to the 3B parameter scale.
Lack of in-depth expertise in specific domains.
Possibility of bias and hallucination.

Ethics Statement

Although ethical considerations have been maximally reflected in the model development process, users should always critically review the results.

📚 Documentation

Model Information Table

Property	Details
License	llama3.2
Training Datasets	CarrotAI/Magpie-Ko-Pro-AIR, CarrotAI/Carrot, CarrotAI/ko-instruction-dataset
Languages	Korean, English
Base Model	meta-llama/Llama-3.2-3B-Instruct
Pipeline Tag	text-generation
New Version	CarrotAI/Llama-3.2-Rabbit-Ko-3B-Instruct-2412

Score

Tasks	Version	Filter	n-shot	Metric	Value	Stderr
gsm8k	3	flexible-extract	5	exact_match	0.6490	0.0131
		strict-match	5	exact_match	0.0023	0.0013
gsm8k-ko	3	flexible-extract	5	exact_match	0.3275	0.0134
		strict-match	5	exact_match	0.2737	0.0134
ifeval	4	none	5	inst_level_loose_acc	0.8058	N/A
		none	5	inst_level_strict_acc	0.7686	N/A
		none	5	prompt_level_loose_acc	0.7320	0.0191
		none	5	prompt_level_strict_acc	0.6858	0.0200

Tasks	Version	Filter	n-shot	Metric	Value	Stderr
haerae	1	none		acc	0.4180	0.0148
		none		acc_norm	0.4180	0.0148
- haerae_general_knowledge	1	none	5	acc	0.3125	0.0350
		none	5	acc_norm	0.3125	0.0350
- haerae_history	1	none	5	acc	0.3404	0.0347
		none	5	acc_norm	0.3404	0.0347
- haerae_loan_word	1	none	5	acc	0.4083	0.0379
		none	5	acc_norm	0.4083	0.0379
- haerae_rare_word	1	none	5	acc	0.4815	0.0249
		none	5	acc_norm	0.4815	0.0249
- haerae_standard_nomenclature	1	none	5	acc	0.4771	0.0405
		none	5	acc_norm	0.4771	0.0405

Tasks	Version	Filter	n-shot	Metric	Value	Stderr
kobest_boolq	1	none	5	acc	0.7664	0.0113
		none	5	f1	0.7662	N/A
kobest_copa	1	none	5	acc	0.5620	0.0157
		none	5	f1	0.5612	N/A
kobest_hellaswag	1	none	5	acc	0.3840	0.0218
		none	5	acc_norm	0.4900	0.0224
		none	5	f1	0.3807	N/A
kobest_sentineg	1	none	5	acc	0.5869	0.0247
		none	5	f1	0.5545	N/A
kobest_wic	1	none	5	acc	0.4952	0.0141
		none	5	f1	0.4000	N/A

📄 License

The license of this model is llama3.2.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご