Hugging Face 推出 Synthetic Data Generator：用自然語言輕鬆構建 AI 訓練資料集
Original: Introducing the Synthetic Data Generator - Build Datasets with Natural Language
Hugging Face launched a brand-new "Synthetic Data Generator" in December 2024 — a web-based, no-code tool designed to allow anyone to…
Hugging Face 發表「Synthetic Data Generator」工具，旨在降低 AI 模型訓練資料集的構建門檻。用戶只需用自然語言描述需求，系統便會利用 distilabel 框架與開源大模型（如 Llama 3.1）自動生成高品質的指令微調（SFT）或偏好對齊（DPO）資料集。生成的資料可直接上傳至 Hugging Face Hub，並支援匯出至 Argilla 進行人工標註與微調。
Hugging Face launched a brand-new "Synthetic Data Generator" in December 2024 — a web-based, no-code tool designed to allow anyone to create high-quality AI training datasets simply by describing what they need in natural language. The release of this tool dramatically lowers the barrier for developers who struggle with a shortage of high-quality data when fine-tuning large language models.
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Hugging Face Blog →Related
Summaries are AI-generated; the original article is authoritative.