Efficient Personalization of Quantized Diffusion Model without Backpropagation

• URL: http://arxiv.org/abs/2503.14868v1
• Date: Wed, 19 Mar 2025 03:45:37 GMT
• Title: Efficient Personalization of Quantized Diffusion Model without Backpropagation
• Authors: Hoigi Seo, Wongi Jeong, Kyungryeol Lee, Se Young Chun,
• Abstract summary: Diffusion models demand extensive computational and memory resources for training, fine-tuning and inference.<n>We address this challenge by quantizing a diffusion model with personalization via Textual Inversion.<n>Our method achieves comparable performance to prior methods in image and text alignment scores for personalizing Stable Diffusion with only forward passes.
• Score: 9.875073051988057
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Diffusion models have shown remarkable performance in image synthesis, but they demand extensive computational and memory resources for training, fine-tuning and inference. Although advanced quantization techniques have successfully minimized memory usage for inference, training and fine-tuning these quantized models still require large memory possibly due to dequantization for accurate computation of gradients and/or backpropagation for gradient-based algorithms. However, memory-efficient fine-tuning is particularly desirable for applications such as personalization that often must be run on edge devices like mobile phones with private data. In this work, we address this challenge by quantizing a diffusion model with personalization via Textual Inversion and by leveraging a zeroth-order optimization on personalization tokens without dequantization so that it does not require gradient and activation storage for backpropagation that consumes considerable memory. Since a gradient estimation using zeroth-order optimization is quite noisy for a single or a few images in personalization, we propose to denoise the estimated gradient by projecting it onto a subspace that is constructed with the past history of the tokens, dubbed Subspace Gradient. In addition, we investigated the influence of text embedding in image generation, leading to our proposed time steps sampling, dubbed Partial Uniform Timestep Sampling for sampling with effective diffusion timesteps. Our method achieves comparable performance to prior methods in image and text alignment scores for personalizing Stable Diffusion with only forward passes while reducing training memory demand up to $8.2\times$.

Related papers

• Efficient Dataset Distillation via Diffusion-Driven Patch Selection for Improved Generalization [34.53986517177061]
  We propose a novel framework to existing diffusion-based distillation methods, leveraging diffusion models for selection rather than generation.<n>Our method starts by predicting noise generated by the diffusion model based on input images and text prompts, then calculates the corresponding loss for each pair.<n>This streamlined framework enables a single-step distillation process, and extensive experiments demonstrate that our approach outperforms state-of-the-art methods across various metrics.
  arXiv  Detail & Related papers  (2024-12-13T08:34:46Z)
• Hollowed Net for On-Device Personalization of Text-to-Image Diffusion Models [51.3915762595891]
  This paper presents an efficient LoRA-based personalization approach for on-device subject-driven generation. Our method, termed Hollowed Net, enhances memory efficiency during fine-tuning by modifying the architecture of a diffusion U-Net.
  arXiv  Detail & Related papers  (2024-11-02T08:42:48Z)
• Fast constrained sampling in pre-trained diffusion models [77.21486516041391]
  Diffusion models have dominated the field of large, generative image models. We propose an algorithm for fast-constrained sampling in large pre-trained diffusion models.
  arXiv  Detail & Related papers  (2024-10-24T14:52:38Z)
• Towards More Accurate Diffusion Model Acceleration with A Timestep Aligner [84.97253871387028]
  A diffusion model, which is formulated to produce an image using thousands of denoising steps, usually suffers from a slow inference speed. We propose a timestep aligner that helps find a more accurate integral direction for a particular interval at the minimum cost. Experiments show that our plug-in design can be trained efficiently and boost the inference performance of various state-of-the-art acceleration methods.
  arXiv  Detail & Related papers  (2023-10-14T02:19:07Z)
• Winner-Take-All Column Row Sampling for Memory Efficient Adaptation of Language Model [89.8764435351222]
  We propose a new family of unbiased estimators called WTA-CRS, for matrix production with reduced variance. Our work provides both theoretical and experimental evidence that, in the context of tuning transformers, our proposed estimators exhibit lower variance compared to existing ones.
  arXiv  Detail & Related papers  (2023-05-24T15:52:08Z)
• Gradient-Free Textual Inversion [34.474779413929426]
  It is natural to question whether we can optimize the textual inversions by only accessing the process inference model inference model. We introduce a emphgradient evolutionary strategy to optimize continuous textual inversion in an iterative evolutionary strategy. Experiments in several applications demonstrate the performance of text-to-image model equipped with our proposed gradient-free method.
  arXiv  Detail & Related papers  (2023-04-12T12:46:27Z)
• Latent Autoregressive Source Separation [5.871054749661012]
  This paper introduces vector-quantized Latent Autoregressive Source Separation (i.e., de-mixing an input signal into its constituent sources) without requiring additional gradient-based optimization or modifications of existing models. Our separation method relies on the Bayesian formulation in which the autoregressive models are the priors, and a discrete (non-parametric) likelihood function is constructed by performing frequency counts over latent sums of addend tokens.
  arXiv  Detail & Related papers  (2023-01-09T17:32:00Z)
• Self-Tuning Stochastic Optimization with Curvature-Aware Gradient Filtering [53.523517926927894]
  We explore the use of exact per-sample Hessian-vector products and gradients to construct self-tuning quadratics. We prove that our model-based procedure converges in noisy gradient setting. This is an interesting step for constructing self-tuning quadratics.
  arXiv  Detail & Related papers  (2020-11-09T22:07:30Z)
• Extrapolation for Large-batch Training in Deep Learning [72.61259487233214]
  We show that a host of variations can be covered in a unified framework that we propose. We prove the convergence of this novel scheme and rigorously evaluate its empirical performance on ResNet, LSTM, and Transformer.
  arXiv  Detail & Related papers  (2020-06-10T08:22:41Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.