Hugging Face released hfdask, a tool that enables synthetic data generation using Dask on Hugging Face Jobs, on September 15, 2026. It is the company's first software update in the area of distributed machine learning since the launch of Hugging Face Jobs in 2025.
Hugging Face reported that hfdask allows users to build a complete CPU → GPU → CPU pipeline for synthetic data generation, measured on a distributed computing framework. That compares with earlier methods that required separate orchestration systems and storage for intermediate boundaries.
hfdask is built on Hugging Face Jobs and targets synthetic data creation for training language models. Availability begins with the release of the tool, initially for developers and data scientists working on NLP tasks.
"hfdask turns a YAML cluster definition into a temporary Dask cluster on Hugging Face Jobs," said Stephen Solka, a contributor at Hanno-Labs. The tool starts one coordinator Job that hosts the Dask scheduler and runs your script, then starts the worker Jobs that contribute CPU and GPU workers.
The announcement follows the release of the Hugging Face Jobs platform, which provides heterogeneous machines and native access to Hub resources. Hugging Face did not say how the tool will be integrated into existing workflows, and it raised the question of whether hfdask will support other synthetic data generation tasks beyond question-and-answer examples.
Source: huggingface
