paper-with-me

홈 › Papers

Scaling Down Text Encoders of Text-to-Image Diffusion Models

2025-03-25 · CVPR 2025 1 · Lifu Wang, Daqing Liu, Xinchen Liu, Xiaodong He

Text encoders in diffusion models have rapidly evolved, transitioning from CLIP to T5-XXL. Although this evolution has significantly enhanced the models' ability to understand complex prompts and generate text, it also leads to a substantial increase in the number of parameters. Despite T5 series encoders being trained on the C4 natural language corpus, which includes a significant amount of non-visual data, diffusion models with T5 encoder do not respond to those non-visual prompts, indicating redundancy in representational power. Therefore, it raises an important question: "Do we really need such a large text encoder?" In pursuit of an answer, we employ vision-based knowledge distillation to train a series of T5 encoder models. To fully inherit its capabilities, we constructed our dataset based on three criteria: image quality, semantic understanding, and text-rendering. Our results demonstrate the scaling down pattern that the distilled T5-base model can generate images of comparable quality to those produced by T5-XXL, while being 50 times smaller in size. This reduction in model size significantly lowers the GPU requirements for running state-of-the-art models such as FLUX and SD3, making high-quality text-to-image generation more accessible.

📄 PDF Abstract BibTeX arXiv:2503.19897

Code (1)

LifuWang-66/DistillT5 공식 구현 jax

Tasks

GPUImage GenerationKnowledge DistillationText to Image GenerationText-to-Image Generation

Methods 이 논문이 사용한 방법론

Gated Linear Unit A Gated Linear Unit, or GLU computes: $$ \mathrm{GLU}(a, b) = a \otimes \sigma(b) $$ It is used in natural language processing architectures, for example the Gated CNN,…
Refunds@Expedia|||How do I get a full refund from Expedia? “How do I get a full refund from Expedia? How do I get a full refund from Expedia? – Call ☎️ +1-(888) 829 (0881) or +1-805-330-4056 or +1-805-330-4056 for Quick Help &…
BPE Byte Pair Encoding, or BPE, is a subword segmentation algorithm that encodes rare and unknown words as sequences of subword units. The intuition is that various word…
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Layer Normalization Unlike batch normalization, Layer Normalization directly estimates the normalization statistics from the summed inputs…
Attention Dropout Attention Dropout is a type of dropout used in attention-based architectures, where elements are randomly dropped out of the…
SentencePiece 설명 없음

Similar Papers 제목 키워드 기반

Dynamic Reflections: Probing Video Representations with Text Alignment

2025-11-04 · Tyler Zhu, Tengda Han, Leonidas Guibas, Viorica Pătrăucean 외 arxiv

The alignment of representations from different modalities has recently been shown to provide insights on the structural similarities and downstream capabilities of different encoders across diverse data types. While sig…

TuringViT: Making SOTA Vision Transformers Accessible to All

2026-06-23 · Qiman Wu, Hanlin Chen, Lyujie Chen, Rui Xin 외 arxiv

Modern VLMs and VLA systems commonly adopt off-the-shelf ViTs such as SigLIP2 as visual encoders, but diverse downstream requirements in latency, temporal modeling, and VLM integration often call for customized SOTA-leve…

Learnings from Scaling Visual Tokenizers for Reconstruction and Generation

2025-01-16 · Philippe Hansen-Estruch, David Yan, Ching-Yao Chung, Orr Zohar 외

Visual tokenization via auto-encoding empowers state-of-the-art image and video generative models by compressing pixels into a latent space. Although scaling Transformer-based generators has been central to recent advanc…

DecoderImage GenerationVideo GenerationVideo Reconstruction

Spatiotemporal Satellite Image Downscaling with Transfer Encoders and Autoregressive Generative Models

2025-12-01 · Yang Xiang, Jingwen Zhong, Yige Yan, Petros Koutrakis 외 arxiv

We present a transfer-learning generative downscaling framework to reconstruct fine resolution satellite images from coarse scale inputs. Our approach combines a lightweight U-Net transfer encoder with a diffusion-based …

Transfer Learning

X-Transfer Attacks: Towards Super Transferable Adversarial Attacks on CLIP

2025-05-08 · Hanxun Huang, Sarah Erfani, Yige Li, Xingjun Ma 외

As Contrastive Language-Image Pre-training (CLIP) models are increasingly adopted for diverse downstream tasks and integrated into large vision-language models (VLMs), their susceptibility to adversarial perturbations ha…