AutoTrain
No-code tool for training and fine-tuning state-of-the-art models on your own data.
About
AutoTrain is Hugging Face's no-code training tool for NLP, computer vision, speech, and tabular tasks. You upload a dataset, pick a task type, and it handles hyperparameter search, training, and evaluation, producing a model ready to push to the Hub.
It is available in two forms: a point-and-click web UI for small datasets, and AutoTrain Advanced, an open-source package and Docker template you can run locally or on cloud hardware for full control. The underlying approach was published at EMNLP 2024, and it integrates with experiment tracking tools like Weights & Biases.
What it can do
- No-code LLM fine-tuning and task training
- LLM, text, vision, speech, and tabular tasks
- AutoTrain Advanced local/cloud training
- W&B experiment tracking integration
Setup
Use the web UI, or pip install autotrain-advanced and run the Docker template in a Hugging Face Space (kept private with attached hardware).
Prerequisites
A Hugging Face account with a write token, and a labeled dataset in a supported format.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 76; Trending score 80/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.