K

Kogpt J 350m

Developed by heegyu
A Korean text generation model based on the GPT-J architecture with 350 million parameters, suitable for various Korean text generation tasks.
Downloads 123
Release Time : 12/28/2022

Model Overview

This model is a Korean text generation model based on the GPT-J architecture, specifically optimized for Korean text generation tasks, capable of producing fluent and coherent Korean text.

Model Features

High-performance Korean text generation
The model excels in various Korean text generation tasks, producing fluent and coherent text.
Large-scale training data
The model was trained on diverse Korean datasets, including news, dialogues, and Wikipedia, totaling approximately 7 billion tokens.
Optimized training environment
The model was trained on TPU V2-8 with efficient training hyperparameters and scheduling strategies.

Model Capabilities

Korean text generation
Dialogue generation
News summarization

Use Cases

Dialogue systems
Daily dialogue generation
The model can generate natural and fluent daily dialogue text.
Example output: '์•ˆ๋…•ํ•˜์„ธ์š”?\n๋„ค.\n์ž~ ์˜ค๋Š˜ ๊ทธ~ ๋ญ~ ๋‚จ๋ถ์ •์ƒํšŒ๋‹ด์—์„œ ์ธ์ œ ๋‚จ๋ถ ๊ด€๊ณ„์™€ ๊ด€๋ จ๋œ ๋ฐœ์–ธ์ด์ฃ ?'
News generation
News summarization
The model can generate news summaries based on prompts.
Example output: '์˜ค๋Š˜ ์ •๋ถ€ ๋ฐœํ‘œ์— ๋”ฐ๋ฅด๋ฉด, gtx-d d ๋…ธ์„ ์„ ์ฐฝ๋ฆ‰๊ณผ ์ˆ˜์„œ์—์„œ ์ถœ๋ฐœํ•˜๋Š” ๋“ฑ ๋‹น์ดˆ ์˜ˆ์ •๋œ ๋…ธ์„ ๋“ค์„ ๋ชจ๋‘ ์ •์ฐจํ•˜๊ธฐ๋กœ ํ–ˆ๋‹ค.'
Featured Recommended AI Models
ยฉ 2025AIbase