Skip to content

Dataset · opencsg

Chinese synthetic dataset for LLM fine-tuning

opencsg/smoltalk-chinese

Over 700,000 synthetic Chinese examples based on SmolTalk for SFT.

What's inside

A dataset from the OpenCSG community, created by analogy with the well-known SmolTalk dataset to improve the quality of Chinese-language language models. It consists entirely of synthetic data in Chinese and numbers over 700,000 records (metadata indicates a range of 10K<n<100K, but the description states over 700,000). Presented in text and table formats.

What you can do with it

  • SFT of language models to improve Chinese generation
  • Training LoRA adapters for dialogue and instruction tasks
Tasks
Text generation
Languages
Chinese
Size
10K–100K rows
Downloads this month
2.8K
Likes
52

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("opencsg/smoltalk-chinese")