SoftCap: Soft-Budget Control for Diffusion Transformer Acceleration
For practitioners deploying diffusion transformers, SoftCap offers a training-free method to improve image quality under a soft compute budget without requiring fixed schedules or hand-tuned thresholds.
SoftCap introduces a training-free control layer for cache-based Diffusion Transformer inference that dynamically adjusts when to execute full model evaluations using a trajectory drift observer and a soft-budget PI controller. On FLUX.1-dev, it improves ImageReward from 0.967 to 0.981 and reduces LPIPS-Full from 0.518 to 0.498 at nearly identical FLOPs compared to SpeCa.
Diffusion Transformers (DiTs) achieve strong visual quality, but their iterative denoising process requires many costly Transformer evaluations. Training-free acceleration methods reduce this cost by caching, forecasting, or verifying intermediate features, yet the runtime decision of when to execute a Full step is often driven by fixed schedules or hand-tuned thresholds. We propose \textbf{SoftCap}, a training-free control layer for cache-based DiT inference. SoftCap couples a Trajectory Drift Observer, which estimates local cache risk from lightweight hidden-state statistics, with a Soft-Budget PI Controller, which adjusts the Full-triggering threshold from realized compute relative to a fixed reference profile. The budget is a soft ceiling: it shapes the threshold but does not require a run to spend a prescribed number of Full evaluations. On FLUX.1-dev, SoftCap improves over SpeCa at a comparable middle-compute operating point, raising ImageReward from 0.967 to 0.981 and reducing LPIPS-Full from 0.518 to 0.498 at nearly identical FLOPs, while target-sweep diagnostics show the intended soft-ceiling behavior as the budget is relaxed.