Model Overview

The ifuseok/sft-solar-10.7b-v1.1 is a 10.7 billion parameter language model built upon the upstage/SOLAR-10.7B-Instruct-v1.0 base architecture. This model has undergone supervised fine-tuning (SFT) using a collection of Korean-centric instruction datasets, enhancing its ability to understand and generate responses in Korean.

Key Capabilities

• Korean Instruction Following: Specialized training on datasets like nlpai-lab/databricks-dolly-15k-ko, kyujinpy/KOR-OpenOrca-Platypus-v3, and heegyu/open-korean-instructions enables robust performance in following Korean instructions.
• Korean Language Understanding: Incorporates datasets such as KETI-AIR/kor_boolq and AIhub Korean-English translation data, improving its comprehension of Korean linguistic nuances.
• Text Generation: Capable of generating coherent and contextually relevant text based on given prompts.

Training Details

The model was fine-tuned using a variety of publicly available Korean datasets, focusing on instruction-response pairs and general Korean language tasks. This targeted training aims to optimize its performance for applications requiring strong Korean language capabilities.

Good For

• Applications requiring a strong Korean instruction-following model.
• Korean-centric chatbots and conversational AI systems.
• Tasks involving Korean text generation, summarization, or question answering.