Overview

Welcome to the OneTrainer wiki! OneTrainer is a customtkinter-based GUI designed to train and work with various diffusion models, primarily built on top of the Diffusers library.

• If you need assistance, the best place is the Discord
• If you have found a bug, problem or have a feature request, search for existing issues first before creating an issue: https://github.com/Nerogar/OneTrainer/issues

Learning

• Getting Started
• The Program - The various tabs of OT
• Info & Guides, and Lessons Learned
• Dev Corner

OneTrainer Requirements

• We require a system that supports PyTorch 2.6.0 or later and a Python version >= 3.10 and < 3.13. For CUDA (machines with Nvidia GPUs) this means a cuda version of 11.8 or higher, for ROCM, 6.2.4 or higher.
• Roughly 7GB free space to install. Training will use more as model files/backups are saved.
• For users looking to do CPU offloading we strongly recommend 64GB, you can get by on 32GB but its likely you will encounter some OOMing due to garbage collection.