kanana-1.5.2.1b-instruct-sft

This is a fine-tuned version of kakaocorp/kanana-1.5-2.1b-instruct-2505.

---

Notice (Required)

Kanana is licensed in accordance with the Kanana License Agreement.
Copyright © KAKAO Corp. All Rights Reserved.

Powered by Kanana

---

Derivative / Modification

This repository contains a fine-tuned derivative of:

Base Model: kakaocorp/kanana-1.5-2.1b-instruct-2505
Modified by: Jiwon Jeong

The model was fine-tuned using supervised fine-tuning (SFT) with custom instruction-following data.

---

Responsible AI

Use of this model must comply with applicable laws and Kakao's Guidelines for Responsible AI:

https://www.kakaocorp.com/page/responsible/detail/guidelinesForResponsibleAI

---

License

This model is distributed under the Kanana License Agreement.

Please refer to the LICENSE file in this repository for the full license text.

---

Model Description

This model is an instruction-following fine-tuned version of Kanana-1.5-2.1B.
It is designed to improve conversational ability and instruction adherence through supervised fine-tuning.

---

Training Details

• Base Model: kakaocorp/kanana-1.5-2.1b-instruct-2505
• Method: Supervised Fine-Tuning (SFT)
• Framework: Axolotl
• Task Type: Instruction Following / Chat

---

Usage

You can load the model using Hugging Face Transformers:

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "jiwon9703/kanana-1.5.2.1b-instruct-sft",
    device_map="auto",
    torch_dtype="auto"
)

tokenizer = AutoTokenizer.from_pretrained("jiwon9703/kanana-1.5.2.1b-instruct-sft")