REVIEW 14 cited by
Accelerating Diffusion Transformers with Token-wise Feature Caching
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
abstract
Diffusion transformers have shown significant effectiveness in both image and video synthesis at the expense of huge computation costs. To address this problem, feature caching methods have been introduced to accelerate diffusion transformers by caching the features in previous timesteps and reusing them in the following timesteps. However, previous caching methods ignore that different tokens exhibit different sensitivities to feature caching, and feature caching on some tokens may lead to 10$\times$ more destruction to the overall generation quality compared with other tokens. In this paper, we introduce token-wise feature caching, allowing us to adaptively select the most suitable tokens for caching, and further enable us to apply different caching ratios to neural layers in different types and depths. Extensive experiments on PixArt-$\alpha$, OpenSora, and DiT demonstrate our effectiveness in both image and video generation with no requirements for training. For instance, 2.36$\times$ and 1.93$\times$ acceleration are achieved on OpenSora and PixArt-$\alpha$ with almost no drop in generation quality.
Forward citations
Cited by 14 Pith papers
-
Token Radius Attention for Efficient Video Generation
Video diffusion transformers can run ~1.5-2x faster with competitive quality by converting each query's attention entropy into a spatially decayed retention radius instead of dense attention.
-
CODA: Algorithm-Hardware Co-design for Edge Video Diffusion via NMP-Enabled Compute-Cache Operator Disaggregation
Disaggregating cache operators from compute and overlapping them across the two classifier-free-guidance branches turns cross-timestep caching into up to 1.80x real end-to-end speedup on edge GPUs when the cache overf...
-
Kaleido: Algorithm-Hardware Co-Design for Video Diffusion Transformers by Exploiting Latent Space Correlations
A channel-wise reuse algorithm plus a reconfigurable systolic accelerator skips redundant vDiT attention and MLP computation, achieving up to 5.9x speedup and 16x energy savings.
-
Phase-Aligned RoPE for Mixed-Resolution Diffusion Transformer
Expressing all RoPE positions on the query's grid ('one attention, one scale') plus a small boundary content-exchange step restores mixed-resolution diffusion generation that naive position interpolation destroys.
-
Timeripple: Accelerating vDiTs by Understanding the Spatio-Temporal Correlations in Latent Space
Timeripple cuts vDiT self-attention compute by up to 85% by reusing partial attention scores of spatially and temporally correlated tokens across channels, with VBench quality essentially unchanged.
-
Forecast then Calibrate: Feature Caching as ODE for Efficient Diffusion Transformers
A training-free predictor-corrector method that accelerates Diffusion Transformers by solving a feature-ODE, achieving large compute reductions with modest quality loss.
-
Yume: An Interactive World Generation Model
A diffusion-based video model generates extendable, keyboard-controlled walkthroughs from a single input image, using quantized camera actions as text prompts.
-
SADA: Stability-guided Adaptive Diffusion Acceleration
SADA accelerates ODE-based generative model sampling by adaptively combining step skipping and token pruning through a stability criterion, giving about 1.8 times speedup with minor fidelity loss.
-
Revisiting Diffusion Models: From Generative Pre-training to One-Step Generation
Fine-tuning a pretrained diffusion model with a GAN objective and most weights frozen yields a one-step generator that matches or beats prior distillation methods on several datasets.
-
CAT Pruning: Cluster-Aware Token Pruning For Text-to-Image Diffusion Models
A token-pruning cache method cuts diffusion model computation by roughly half while keeping image quality, using noise magnitude, spatial clustering, and selection balance.
-
OmniCache: Multidimensional Hierarchical Feature Caching For Diffusion Models
Training-free hierarchical feature caching across token, frame, block, and layer axes cuts diffusion inference latency up to 35% while preserving quality better than averaging-based token merging.
-
BeCARE: Budgeted Cache Refresh for Diffusion Transformer Acceleration
SoftCap introduces a Trajectory Drift Observer and Soft-Budget PI Controller for adaptive full-step triggering in DiT caching, reporting better ImageReward and LPIPS than SpeCa at similar FLOPs on FLUX.1-dev.
-
Accelerating Diffusion Transformer via Error-Optimized Cache
A precomputed trend correction injected into cached diffusion transformer blocks reduces FID by up to 28.8% at heavy caching with negligible added compute.
-
SkipVAR: Accelerating Visual Autoregressive Modeling via Adaptive Frequency-Aware Skipping
SkipVAR selects, per sample, between step skipping and unconditional branch replacement using handcrafted frequency features and a trained logistic regression, to accelerate visual autoregressive generation.
Discussion (0). Continue with ORCID to comment.