Ko-Llama

When this dataset fits

Ko-Llama is a Korean Instruction Tuning Dataset from Hugging Face. It fits instruction tuning, not long-document pretraining. License label Apache 2.0 still needs review. Size text is “대용량”, copied from the host.

Open 1 catalog excerpts →

Task / category

Text · Text · Llama, SFT, instruction

Language

Korean

Format · Size

Text · Large (80K)

License

Apache 2.0(Catalog hint — re-check original) · License document

Source & original link

Hugging Face · URL not verified

Verification status

Source: Unknown · License: Needs review · Last checked: 2026-09-03

Ko-Llama: Korean Instruction Tuning Dataset from Hugging Face. License Apache 2.0. Open the original link to download.

Ko-Llama (2023). Hugging Face. license label: Apache 2.0 catalog commercial mark: allowed Original URL is not stored on this catalog row. GearDel catalog: https://geardel.com/en/datasets/ko-llama-instruct-v1. Last checked: 2026-09-03. Author names are not stored in the GearDel catalog and are not invented here.

Differences below use catalog fields only: license, commercial mark, provider, size text, and year.

• Korean Instruction Tuning DatasetKorean Instruction Tuning Dataset · Hugging FaceViewLicense labels differ (Apache 2.0 vs MIT).
• QARVKorean Instruction Tuning Dataset · Hugging FaceViewLicense labels differ (Apache 2.0 vs MIT). Size text differs (Large vs Medium). Those strings are copied from the hosts. Catalog years differ (2023 vs 2024).
• Korean Instruction Tuning DatasetKorean Instruction Tuning Dataset · Hugging FaceViewLicense labels differ (Apache 2.0 vs MIT).
• QGKorean Instruction Tuning Dataset · Hugging FaceViewSize text differs (Large vs Small). Those strings are copied from the hosts. Catalog years differ (2023 vs 2022).

EXCERPTS

Catalog excerpts

1 stored samples. GearDel does not host the original files.

Open this row in Try samples →

가장 대표적인 봄철 제철 과일 딸기를 주제로 짧은 4행시를 지어 주세요.