Skip to content

Dataset · Aratako

Japanese and English synthetic coding dataset

Aratako/Synthetic-JP-EN-Coding-Dataset-801k

801 thousand synthetic instruction-response pairs for SFT training of programming models in Japanese and English.

What's inside

Contains 801,262 SFT examples for programming: 627,413 in English and 173,849 in Japanese. The base was a 69-thousand-record dataset created via the Magpie method, which was then expanded using an Evol-Instruct-like approach. Open LLMs, including Nemotron-4-340B, Phi-3-medium, Mixtral-8x22B, and calm3-22b-chat, participated in generating the source instructions.

What you can do with it

  • SFT fine-tuning of LLMs for code generation in Japanese and English
  • Training multilingual LoRA adapters for developer assistants
Tasks
Text generation
Languages
Japanese
Size
100K–1M rows
Downloads this month
291
Likes
18

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Aratako/Synthetic-JP-EN-Coding-Dataset-801k")