--- license: cc datasets: - Amber-River/Pixiv-2.6M - KBlueLeaf/danbooru2023-webp-4Mpixel language: - en pipeline_tag: text-to-image library_name: diffusers tags: - text-generation-inference - anime --- # HDM-xut-340M-Anime ***World's smallest, cheapest anime-style T2I base*** [](https://github.com/KohakuBlueleaf/HDM) [-2ea44f)](https://github.com/KohakuBlueleaf/HDM-ext) [-HDM-2ea44f)](https://huggingface.co/KBlueLeaf/HDM-xut-340M-anime) [](https://github.com/KohakuBlueleaf/HDM/blob/main/TechReport.md)  ## Introduction HDM(Home made Diffusion Model) is a project to investigate specialized training recipe/scheme for "pretraining T2I model at home" which require the training setup should be exectuable on customer level hardware or cheap enough second handed server hardware. Under this constraint, we introduce a new transformer backbone designed for multi-modal (for example, text-to-image) generative model called "XUT" (Cross-U-Transformer). With minimalized arch design with [TREAD](https://arxiv.org/abs/2501.04765) technique, we can achieve usable performan with computes that cost 650USD at most. (based on pricing from vast.ai) ## Gallery