Memory-Efficient Fine-Tuning Diffusion Transformers via Dynamic Patch Sampling and Block Skipping

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Park, Sunghyun, Kim, Jeongho, Park, Hyoungwoo, Das, Debasmit, Yun, Sungrack, Hayat, Munawar, Choo, Jaegul, Porikli, Fatih, Choi, Seokeon
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912977335091200
author Park, Sunghyun
Kim, Jeongho
Park, Hyoungwoo
Das, Debasmit
Yun, Sungrack
Hayat, Munawar
Choo, Jaegul
Porikli, Fatih
Choi, Seokeon
author_facet Park, Sunghyun
Kim, Jeongho
Park, Hyoungwoo
Das, Debasmit
Yun, Sungrack
Hayat, Munawar
Choo, Jaegul
Porikli, Fatih
Choi, Seokeon
contents Diffusion Transformers (DiTs) have significantly enhanced text-to-image (T2I) generation quality, enabling high-quality personalized content creation. However, fine-tuning these models requires substantial computational complexity and memory, limiting practical deployment under resource constraints. To tackle these challenges, we propose a memory-efficient fine-tuning framework called DiT-BlockSkip, integrating timestep-aware dynamic patch sampling and block skipping by precomputing residual features. Our dynamic patch sampling strategy adjusts patch sizes based on the diffusion timestep, then resizes the cropped patches to a fixed lower resolution. This approach reduces forward & backward memory usage while allowing the model to capture global structures at higher timesteps and fine-grained details at lower timesteps. The block skipping mechanism selectively fine-tunes essential transformer blocks and precomputes residual features for the skipped blocks, significantly reducing training memory. To identify vital blocks for personalization, we introduce a block selection strategy based on cross-attention masking. Evaluations demonstrate that our approach achieves competitive personalization performance qualitatively and quantitatively, while reducing memory usage substantially, moving toward on-device feasibility (e.g., smartphones, IoT devices) for large-scale diffusion transformers.
format Preprint
id arxiv_https___arxiv_org_abs_2603_20755
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Memory-Efficient Fine-Tuning Diffusion Transformers via Dynamic Patch Sampling and Block Skipping
Park, Sunghyun
Kim, Jeongho
Park, Hyoungwoo
Das, Debasmit
Yun, Sungrack
Hayat, Munawar
Choo, Jaegul
Porikli, Fatih
Choi, Seokeon
Computer Vision and Pattern Recognition
Artificial Intelligence
Diffusion Transformers (DiTs) have significantly enhanced text-to-image (T2I) generation quality, enabling high-quality personalized content creation. However, fine-tuning these models requires substantial computational complexity and memory, limiting practical deployment under resource constraints. To tackle these challenges, we propose a memory-efficient fine-tuning framework called DiT-BlockSkip, integrating timestep-aware dynamic patch sampling and block skipping by precomputing residual features. Our dynamic patch sampling strategy adjusts patch sizes based on the diffusion timestep, then resizes the cropped patches to a fixed lower resolution. This approach reduces forward & backward memory usage while allowing the model to capture global structures at higher timesteps and fine-grained details at lower timesteps. The block skipping mechanism selectively fine-tunes essential transformer blocks and precomputes residual features for the skipped blocks, significantly reducing training memory. To identify vital blocks for personalization, we introduce a block selection strategy based on cross-attention masking. Evaluations demonstrate that our approach achieves competitive personalization performance qualitatively and quantitatively, while reducing memory usage substantially, moving toward on-device feasibility (e.g., smartphones, IoT devices) for large-scale diffusion transformers.
title Memory-Efficient Fine-Tuning Diffusion Transformers via Dynamic Patch Sampling and Block Skipping
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2603.20755