Hugging Face shows its ML Intern agent training small custom models for a few dollars
In a Hugging Face blog post, staff describe using the ML Intern agent in HuggingChat to plan, train, evaluate and publish small models. One 0.8B model cost about $16 in compute.
A Hugging Face blog post published October 8 by Yuvraj Sharma and Abubakar Abid walks through building models with ML Intern, an agent that can be switched on in HuggingChat.
Sharma writes that he asked the agent for a small version of the 9B prompt rewriter that ships with Qwen-Image 2.1 and received a 0.8B model that runs on a CPU, returns valid output 99.7% of the time and uses about a quarter of the teacher model's tokens. He says total compute, including having the 9B model label 8,797 examples, came to $16.
According to the post, the agent plans the work, asks for a budget before spending, runs a small test, then trains, evaluates and publishes on Hugging Face hardware. The author says he built five more models the same way and has published his prompts on GitHub.