Salvato in:
Dettagli Bibliografici
Autore principale: Yeh, Shih-Ying
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2509.06068
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914027077107712
author Yeh, Shih-Ying
author_facet Yeh, Shih-Ying
contents We introduce Home-made Diffusion Model (HDM), an efficient yet powerful text-to-image diffusion model optimized for training (and inferring) on consumer-grade hardware. HDM achieves competitive 1024x1024 generation quality while maintaining a remarkably low training cost of $535-620 using four RTX5090 GPUs, representing a significant reduction in computational requirements compared to traditional approaches. Our key contributions include: (1) Cross-U-Transformer (XUT), a novel U-shape transformer, Cross-U-Transformer (XUT), that employs cross-attention for skip connections, providing superior feature integration that leads to remarkable compositional consistency; (2) a comprehensive training recipe that incorporates TREAD acceleration, a novel shifted square crop strategy for efficient arbitrary aspect-ratio training, and progressive resolution scaling; and (3) an empirical demonstration that smaller models (343M parameters) with carefully crafted architectures can achieve high-quality results and emergent capabilities, such as intuitive camera control. Our work provides an alternative paradigm of scaling, demonstrating a viable path toward democratizing high-quality text-to-image generation for individual researchers and smaller organizations with limited computational resources.
format Preprint
id arxiv_https___arxiv_org_abs_2509_06068
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Home-made Diffusion Model from Scratch to Hatch
Yeh, Shih-Ying
Computer Vision and Pattern Recognition
We introduce Home-made Diffusion Model (HDM), an efficient yet powerful text-to-image diffusion model optimized for training (and inferring) on consumer-grade hardware. HDM achieves competitive 1024x1024 generation quality while maintaining a remarkably low training cost of $535-620 using four RTX5090 GPUs, representing a significant reduction in computational requirements compared to traditional approaches. Our key contributions include: (1) Cross-U-Transformer (XUT), a novel U-shape transformer, Cross-U-Transformer (XUT), that employs cross-attention for skip connections, providing superior feature integration that leads to remarkable compositional consistency; (2) a comprehensive training recipe that incorporates TREAD acceleration, a novel shifted square crop strategy for efficient arbitrary aspect-ratio training, and progressive resolution scaling; and (3) an empirical demonstration that smaller models (343M parameters) with carefully crafted architectures can achieve high-quality results and emergent capabilities, such as intuitive camera control. Our work provides an alternative paradigm of scaling, demonstrating a viable path toward democratizing high-quality text-to-image generation for individual researchers and smaller organizations with limited computational resources.
title Home-made Diffusion Model from Scratch to Hatch
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.06068