SimpleTuner
A general-purpose fine-tuning toolkit for image, video, and audio diffusion models, built for simplicity and easily understood code.
SimpleTuner is a fine-tuning toolkit for diffusion models covering images, video, and audio, built as a shared academic project that favors code that is easy to read and modify over maximal feature surface. It aims to work well by default for most datasets, from a handful of images to large collections, while only adding techniques with demonstrated benefit.
It provides a web UI for managing a training run end to end, alongside a traditional command-line quickstart, and supports multi-GPU training plus memory-saving techniques such as DeepSpeed optimizer-state offload and FSDP2 sharding with context parallelism for very large datasets or memory-constrained hardware. No data is sent to third parties except through explicitly opted-in flags for reporting or webhooks.
It is written in Python and released under the AGPL-3.0 license, with a community on Discord.