HarmoniCa: Harmonizing Training and Inference for Better Feature Caching in Diffusion Transformer Acceleration

• URL: http://arxiv.org/abs/2410.01723v6
• Date: Sat, 31 May 2025 17:58:24 GMT
• Title: HarmoniCa: Harmonizing Training and Inference for Better Feature Caching in Diffusion Transformer Acceleration
• Authors: Yushi Huang, Zining Wang, Ruihao Gong, Jing Liu, Xinjie Zhang, Jinyang Guo, Xianglong Liu, Jun Zhang,
• Abstract summary: We develop a novel learning-based caching framework dubbed HarmoniCa.<n>It incorporates Step-Wise Denoising Training (SDT) to ensure the continuity of the denoising process.<n>Our framework achieves over $40%$ latency reduction (i.e., $2.07times$ theoretical speedup) and improved performance on PixArt-$alpha$.
• Score: 31.982294870690925
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Diffusion Transformers (DiTs) excel in generative tasks but face practical deployment challenges due to high inference costs. Feature caching, which stores and retrieves redundant computations, offers the potential for acceleration. Existing learning-based caching, though adaptive, overlooks the impact of the prior timestep. It also suffers from misaligned objectives--aligned predicted noise vs. high-quality images--between training and inference. These two discrepancies compromise both performance and efficiency. To this end, we harmonize training and inference with a novel learning-based caching framework dubbed HarmoniCa. It first incorporates Step-Wise Denoising Training (SDT) to ensure the continuity of the denoising process, where prior steps can be leveraged. In addition, an Image Error Proxy-Guided Objective (IEPO) is applied to balance image quality against cache utilization through an efficient proxy to approximate the image error. Extensive experiments across $8$ models, $4$ samplers, and resolutions from $256\times256$ to $2K$ demonstrate superior performance and speedup of our framework. For instance, it achieves over $40\%$ latency reduction (i.e., $2.07\times$ theoretical speedup) and improved performance on PixArt-$\alpha$. Remarkably, our image-free approach reduces training time by $25\%$ compared with the previous method. Our code is available at https://github.com/ModelTC/HarmoniCa.

Related papers

• Steering One-Step Diffusion Model with Fidelity-Rich Decoder for Fast Image Compression [36.10674664089876]
  SODEC is a novel single-step diffusion-based image compression model.<n>It improves fidelity resulting from over-reliance on generative priors.<n>It significantly outperforms existing methods, achieving superior rate-distortion-perception performance.
  arXiv  Detail & Related papers  (2025-08-07T02:24:03Z)
• Representation Entanglement for Generation:Training Diffusion Transformers Is Much Easier Than You Think [56.539823627694304]
  REPA and its variants effectively mitigate training challenges in diffusion models by incorporating external visual representations from pretrained models.<n>We argue that the external alignment, which is absent during the entire denoising inference process, falls short of fully harnessing the potential of discriminative representations.<n>We propose Representation Entanglement for Generation (REG), which entangles low-level image latents with a single high-level class token from pretrained foundation models for denoising.
  arXiv  Detail & Related papers  (2025-07-02T08:29:18Z)
• Step-level Reward for Free in RL-based T2I Diffusion Model Fine-tuning [23.02076024811612]
  Recent advances in text-to-image (T2I) diffusion model fine-tuning leverage reinforcement learning (RL) to align generated images with learnable reward functions.<n>Existing approaches reformulate denoising as a Markov decision process for RL-driven optimization.<n>We propose a credit assignment framework that dynamically distributes dense rewards across denoising steps.
  arXiv  Detail & Related papers  (2025-05-25T15:43:54Z)
• AB-Cache: Training-Free Acceleration of Diffusion Models via Adams-Bashforth Cached Feature Reuse [19.13826316844611]
  Diffusion models have demonstrated remarkable success in generative tasks, yet their iterative denoising process results in slow inference. We provide a theoretical understanding by analyzing the denoising process through the second-order Adams-Bashforth method. We propose a novel caching-based acceleration approach for diffusion models, instead of directly reusing cached results.
  arXiv  Detail & Related papers  (2025-04-13T08:29:58Z)
• Towards Better Alignment: Training Diffusion Models with Reinforcement Learning Against Sparse Rewards [52.90573877727541]
  reinforcement learning (RL) has been considered for diffusion model fine-tuning. RL's effectiveness is limited by the challenge of sparse reward. $textB2text-DiffuRL$ is compatible with existing optimization algorithms.
  arXiv  Detail & Related papers  (2025-03-14T09:45:19Z)
• Fast constrained sampling in pre-trained diffusion models [77.21486516041391]
  We propose an algorithm that enables fast and high-quality generation under arbitrary constraints. During inference, we can interchange between gradient updates computed on the noisy image and updates computed on the final, clean image. Our approach produces results that rival or surpass the state-of-the-art training-free inference approaches.
  arXiv  Detail & Related papers  (2024-10-24T14:52:38Z)
• FreCaS: Efficient Higher-Resolution Image Generation via Frequency-aware Cascaded Sampling [13.275724439963188]
  FreCaS decomposes the sampling process into cascaded stages with gradually increased resolutions. FreCaS significantly outperforms state-of-the-art methods in image quality and generation speed.
  arXiv  Detail & Related papers  (2024-10-24T03:56:44Z)
• ExpertFlow: Optimized Expert Activation and Token Allocation for Efficient Mixture-of-Experts Inference [41.41316718220569]
  ExpertFlow is designed to enhance inference efficiency by accommodating flexible routing and enabling efficient expert scheduling between CPU and GPU. Our experiments demonstrate that ExpertFlow achieves up to 93.72% GPU memory savings and enhances inference speed by 2 to 10 times compared to baseline methods.
  arXiv  Detail & Related papers  (2024-10-23T15:24:54Z)
• Temporal Feature Matters: A Framework for Diffusion Model Quantization [105.3033493564844]
  Diffusion models rely on the time-step for the multi-round denoising. We introduce a novel quantization framework that includes three strategies. This framework preserves most of the temporal information and ensures high-quality end-to-end generation.
  arXiv  Detail & Related papers  (2024-07-28T17:46:15Z)
• Learning-to-Cache: Accelerating Diffusion Transformer via Layer Caching [56.286064975443026]
  We make an interesting and somehow surprising observation: the computation of a large proportion of layers in the diffusion transformer, through a caching mechanism, can be readily removed even without updating the model parameters. We introduce a novel scheme, named Learningto-Cache (L2C), that learns to conduct caching in a dynamic manner for diffusion transformers. Experimental results show that L2C largely outperforms samplers such as DDIM and DPM-r, alongside prior cache-based methods at the same inference speed.
  arXiv  Detail & Related papers  (2024-06-03T18:49:57Z)
• Adaptive Rentention & Correction for Continual Learning [114.5656325514408]
  A common problem in continual learning is the classification layer's bias towards the most recent task. We name our approach Adaptive Retention & Correction (ARC) ARC achieves an average performance increase of 2.7% and 2.6% on the CIFAR-100 and Imagenet-R datasets.
  arXiv  Detail & Related papers  (2024-05-23T08:43:09Z)
• Implicit Image-to-Image Schrodinger Bridge for Image Restoration [13.138398298354113]
  We introduce the Implicit Image-to-Image Schr"odinger Bridge (I$3$SB) to further accelerate the generative process of I$2$SB. I$3$SB restructures the generative process into a non-Markovian framework by incorporating the initial corrupted image at each generative step. Compared to I$2$SB, I$3$SB achieves the same perceptual quality with fewer generative steps, while maintaining or improving fidelity to the ground truth.
  arXiv  Detail & Related papers  (2024-03-10T03:22:57Z)
• Retraining-free Model Quantization via One-Shot Weight-Coupling Learning [41.299675080384]
  Mixed-precision quantization (MPQ) is advocated to compress the model effectively by allocating heterogeneous bit-width for layers. MPQ is typically organized into a searching-retraining two-stage process. In this paper, we devise a one-shot training-searching paradigm for mixed-precision model compression.
  arXiv  Detail & Related papers  (2024-01-03T05:26:57Z)
• DeepCache: Accelerating Diffusion Models for Free [65.02607075556742]
  DeepCache is a training-free paradigm that accelerates diffusion models from the perspective of model architecture. DeepCache capitalizes on the inherent temporal redundancy observed in the sequential denoising steps of diffusion models. Under the same throughput, DeepCache effectively achieves comparable or even marginally improved results with DDIM or PLMS.
  arXiv  Detail & Related papers  (2023-12-01T17:01:06Z)
• ACT-Diffusion: Efficient Adversarial Consistency Training for One-step Diffusion Models [59.90959789767886]
  We show that optimizing consistency training loss minimizes the Wasserstein distance between target and generated distributions. By incorporating a discriminator into the consistency training framework, our method achieves improved FID scores on CIFAR10 and ImageNet 64$times$64 and LSUN Cat 256$times$256 datasets.
  arXiv  Detail & Related papers  (2023-11-23T16:49:06Z)
• On the Effectiveness of LayerNorm Tuning for Continual Learning in Vision Transformers [47.77328392236625]
  State-of-the-art rehearsal-free continual learning methods exploit the peculiarities of Vision Transformers to learn task-specific prompts. We introduce a two-stage training procedure, where we first optimize the task-specific parameters and then train the classifier with the same selection procedure of the inference time. Our method achieves results that are either superior or on par with the state of the art while being computationally cheaper.
  arXiv  Detail & Related papers  (2023-08-18T15:11:16Z)
• Adaptive Cross Batch Normalization for Metric Learning [75.91093210956116]
  Metric learning is a fundamental problem in computer vision. We show that it is equally important to ensure that the accumulated embeddings are up to date. In particular, it is necessary to circumvent the representational drift between the accumulated embeddings and the feature embeddings at the current training iteration.
  arXiv  Detail & Related papers  (2023-03-30T03:22:52Z)
• Efficient Diffusion Training via Min-SNR Weighting Strategy [78.5801305960993]
  We treat the diffusion training as a multi-task learning problem and introduce a simple yet effective approach referred to as Min-SNR-$gamma$. Our results demonstrate a significant improvement in converging speed, 3.4$times$ faster than previous weighting strategies. It is also more effective, achieving a new record FID score of 2.06 on the ImageNet $256times256$ benchmark using smaller architectures than that employed in previous state-of-the-art.
  arXiv  Detail & Related papers  (2023-03-16T17:59:56Z)
• Online Convolutional Re-parameterization [51.97831675242173]
  We present online convolutional re- parameterization (OREPA), a two-stage pipeline, aiming to reduce the huge training overhead by squeezing the complex training-time block into a single convolution. Compared with the state-of-the-art re-param models, OREPA is able to save the training-time memory cost by about 70% and accelerate the training speed by around 2x. We also conduct experiments on object detection and semantic segmentation and show consistent improvements on the downstream tasks.
  arXiv  Detail & Related papers  (2022-04-02T09:50:19Z)
• Accelerating Deep Learning Classification with Error-controlled Approximate-key Caching [72.50506500576746]
  We propose a novel caching paradigm, that we named approximate-key caching. While approximate cache hits alleviate DL inference workload and increase the system throughput, they however introduce an approximation error. We analytically model our caching system performance for classic LRU and ideal caches, we perform a trace-driven evaluation of the expected performance, and we compare the benefits of our proposed approach with the state-of-the-art similarity caching.
  arXiv  Detail & Related papers  (2021-12-13T13:49:11Z)
• 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)
• BERT Loses Patience: Fast and Robust Inference with Early Exit [91.26199404912019]
  We propose Patience-based Early Exit as a plug-and-play technique to improve the efficiency and robustness of a pretrained language model. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers.
  arXiv  Detail & Related papers  (2020-06-07T13:38:32Z)

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.