Chipmunk: Training-Free Acceleration of Diffusion Transformers with Dynamic Column-Sparse Deltas

• URL: http://arxiv.org/abs/2506.03275v1
• Date: Tue, 03 Jun 2025 18:03:32 GMT
• Title: Chipmunk: Training-Free Acceleration of Diffusion Transformers with Dynamic Column-Sparse Deltas
• Authors: Austin Silveria, Soham V. Govande, Daniel Y. Fu,
• Abstract summary: A common observation is that DiT latent noise vectors change slowly across inference steps, which suggests that the DiT compute may be redundant across steps.<n>In this paper, we aim to speed up inference by reducing this redundancy, without additional training.
• Score: 4.7236887095921
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Diffusion Transformers (DiTs) have achieved state-of-the-art performance in high-quality image and video generation but incur substantial compute cost at inference. A common observation is that DiT latent noise vectors change slowly across inference steps, which suggests that the DiT compute may be redundant across steps. In this paper, we aim to speed up inference by reducing this redundancy, without additional training. We first study how activations change between steps in two state-of-the-art open-source DiTs. We find that just 5-25% of the values in attention and MLP explain 70-90% of the change in activations across steps. This finding motivates our approach, Chipmunk, which uses dynamic sparsity at inference time to recompute only the fastest-changing intermediate activations, while caching the rest. Dynamic sparsity introduces two systems challenges: (1) sparse attention and MLP operations tend to underutilize GPU tensor cores; and (2) computing dynamic sparsity patterns at runtime and caching activations both introduce overhead. To address these challenges, Chipmunk first uses a voxel-based reordering of input tokens to introduce column-wise sparsity. We implement column-sparse kernels utilizing efficient sparse gathers from global to shared GPU memory, achieving a 9.3x speedup at 93% sparsity compared to highly-optimized dense baselines. Second, Chipmunk overlaps the computation of sparsity patterns and cache updates with other parts of the computation (e.g., second layer of the MLP) to hide the extra latency. Chipmunk achieves up to 2.16x speedup on HunyuanVideo and 1.41x on FLUX.1-dev without compromising generation quality. Furthermore, we show that Chipmunk can be stacked on top of full step caching, achieving a 3.72x speedup on HunyuanVideo, a 2.67x speedup on WAN2.1, and a 2.25x speedup on FLUX.1-dev with minimal quality impact.

Related papers

• Spark Transformer: Reactivating Sparsity in FFN and Attention [63.20677098823873]
  We introduce Spark Transformer, a novel architecture that achieves a high level of activation sparsity in both FFN and the attention mechanism.<n>This sparsity translates to a 2.5x reduction in FLOPs, leading to decoding wall-time speedups of up to 1.79x on CPU and 1.40x on GPU.
  arXiv  Detail & Related papers  (2025-06-07T03:51:13Z)
• Sparse-vDiT: Unleashing the Power of Sparse Attention to Accelerate Video Diffusion Transformers [24.105473321347894]
  We propose Sparse-vDiT, a sparsity acceleration framework for Video Diffusion Transformer (vDiT)<n>We show that Sparse-vDiT achieves 2.09$times$, 2.38$times$, and 1.67$times$ theoretical FLOP reduction, and actual inference speedups of 1.76$times$, 1.85$times$, and 1.58$times$, respectively.<n>Our work demonstrates that latent structural sparsity in vDiTs can be systematically exploited for long video synthesis.
  arXiv  Detail & Related papers  (2025-06-03T16:42:37Z)
• dKV-Cache: The Cache for Diffusion Language Models [53.85291644298835]
  Diffusion Language Models (DLMs) have been seen as a promising competitor for autoregressive language models.<n>We propose a KV-cache-like mechanism, delayed KV-Cache, for the denoising process of DLMs.<n>Our approach is motivated by the observation that different tokens have distinct representation dynamics throughout the diffusion process.
  arXiv  Detail & Related papers  (2025-05-21T17:32:10Z)
• High-Frequency Prior-Driven Adaptive Masking for Accelerating Image Super-Resolution [87.56382172827526]
  High-frequency regions are most critical for reconstruction.<n>We propose a training-free adaptive masking module for acceleration.<n>Our method reduces FLOPs by 24--43% for state-of-the-art models.
  arXiv  Detail & Related papers  (2025-05-11T13:18:03Z)
• Magic 1-For-1: Generating One Minute Video Clips within One Minute [53.07214657235465]
  We present Magic 1-For-1 (Magic141), an efficient video generation model with optimized memory consumption and inference latency.<n>By applying a test time sliding window, we are able to generate a minute-long video within one minute with significantly improved visual quality and motion dynamics.
  arXiv  Detail & Related papers  (2025-02-11T16:58:15Z)
• Chameleon: An Efficient FHE Scheme Switching Acceleration on GPUs [17.536473118470774]
  homomorphic encryption (FHE) enables direct computation on encrypted data. Existing efforts primarily focus on single-class FHE schemes, which fail to meet the diverse requirements of data types and functions. We present an efficient GPU-based FHE switching acceleration scheme named Chameleon.
  arXiv  Detail & Related papers  (2024-10-08T11:37:49Z)
• 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)
• Recurrent Drafter for Fast Speculative Decoding in Large Language Models [18.342742904042673]
  We present Recurrent Drafter, an advanced speculative decoding approach.<n>It achieves state-of-the-art speedup for large language models (LLMs) inference.
  arXiv  Detail & Related papers  (2024-03-14T23:40:56Z)
• 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)
• SparseViT: Revisiting Activation Sparsity for Efficient High-Resolution Vision Transformer [42.440822037774645]
  We introduce SparseViT that revisits activation sparsity for recent window-based vision transformers (ViTs) SparseViT achieves speedups of 1.5x, 1.4x, and 1.3x compared to its dense counterpart in monocular 3D object detection, 2D instance segmentation, and 2D semantic segmentation.
  arXiv  Detail & Related papers  (2023-03-30T17:59:58Z)

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.