AutoTrain

AutoTrain supports model training on your own hardware through AutoTrain Advanced, a no-code application for people fine-tuning LLMs or building models from their data. Advanced can run locally or on cloud infrastructure. The project is no longer maintained, and it won't receive bug fixes or new features.

LLM tasks include supervised fine-tuning, DPO and ORPO preference training, and reward-model fine-tuning. Advanced also supports text classification and regression, token classification, sequence-to-sequence tasks and extractive question answering. Image tasks include classification and scoring or regression.

The software is free and open source under Apache 2.0. It has a user interface and command-line workflows, and can run on Hugging Face Spaces or with its UI on Colab. Its training stack uses PyTorch, with NVIDIA CUDA in the documented local setup. A local training example fine-tunes HuggingFaceTB/SmolLM2-1.7B-Instruct using quantization and parameter-efficient fine-tuning.

The hosted Hugging Face AutoTrain service automatically selects models for uploaded data, trains and evaluates them, and makes the results available on the Hub. Hosted training uses paid compute resources. Data uploaded to that service stays on Hugging Face's servers, private to your account, and transfers use encryption. The hosted service accepts CSV, TSV, JSON or ZIP files depending on the task.