Zhipeng Di

Suyi Li, Lingyun Yang, Xiaoxiao Jiang et al.

Text-to-image (T2I) generation using diffusion models has become a blockbuster service in today's AI cloud. A production T2I service typically involves a serving workflow where a base diffusion model is augmented with various "add-on" modules, notably ControlNet and LoRA, to enhance image generation control. Compared to serving the base model alone, these add-on modules introduce significant loading and computational overhead, resulting in increased latency. In this paper, we present SwiftDiffusion, a system that efficiently serves a T2I workflow through a holistic approach. SwiftDiffusion decouples ControNet from the base model and deploys it as a separate, independently scaled service on dedicated GPUs, enabling ControlNet caching, parallelization, and sharing. To mitigate the high loading overhead of LoRA serving, SwiftDiffusion employs a bounded asynchronous LoRA loading (BAL) technique, allowing LoRA loading to overlap with the initial base model execution by up to k steps without compromising image quality. Furthermore, SwiftDiffusion optimizes base model execution with a novel latent parallelism technique. Collectively, these designs enable SwiftDiffusion to outperform the state-of-the-art T2I serving systems, achieving up to 7.8x latency reduction and 1.6x throughput improvement in serving SDXL models on H800 GPUs, without sacrificing image quality.

Hong Zhang, Zhongjie Duan, Xingjun Wang et al.

Unified multimodal generative models aim to integrate image understanding and generation abilities, offering significant advantages in harnessing multimodal corpora, particularly interleaved text-image data. However, existing unified models exhibit limitations in image synthesis quality, autoregressive error accumulation, and image editing capability. In this work, we propose Nexus-Gen, a novel architecture that unifies image understanding, generation, and editing tasks in a shared image embedding space. This shared space serves as a bridge for the autoregressive and diffusion models, which seamlessly integrates their complementary strengths in cross-modal modeling. To mitigate the severe error accumulation during autoregressive embedding prediction, we propose a novel prefilled autoregression strategy that aligns training-inference dynamics by prefilling input sequences with learnable embeddings. After multi-stage and multi-task training on our constructed large-scale dataset with 26.3 million samples, Nexus-Gen achieves state-of-the-art performance on the evaluation benchmarks spanning image understanding, generation and editing tasks. All models, datasets, and source codes are released in https://github.com/modelscope/Nexus-Gen to facilitate further advancements across the field.

Lingyun Yang, Suyi Li, Tianyu Feng et al.

Text-to-image generation executes a diffusion workflow comprising multiple models centered on a base diffusion model. Existing serving systems treat each workflow as an opaque monolith, provisioning, placing, and scaling all constituent models together, which obscures internal dataflow, prevents model sharing, and enforces coarse-grained resource management. In this paper, we make a case for micro-serving diffusion workflows with LegoDiffusion, a system that decomposes a workflow into loosely coupled model-execution nodes that can be independently managed and scheduled. By explicitly managing individual model inference, LegoDiffusion unlocks cluster-scale optimizations, including per-model scaling, model sharing, and adaptive model parallelism. Collectively, LegoDiffusion outperforms existing diffusion workflow serving systems, sustaining up to 3x higher request rates and tolerating up to 8x higher burst traffic.

Xiaoxiao Jiang, Suyi Li, Lingyun Yang et al.

Generative image editing using diffusion models has become a prevalent application in today's AI cloud services. In production environments, image editing typically involves a mask that specifies the regions of an image template to be edited. The use of masks provides direct control over the editing process and introduces sparsity in the model inference. In this paper, we present InstGenIE, a system that efficiently serves image editing requests. The key insight behind InstGenIE is that image editing only modifies the masked regions of image templates while preserving the original content in the unmasked areas. Driven by this insight, InstGenIE judiciously skips redundant computations associated with the unmasked areas by reusing cached intermediate activations from previous inferences. To mitigate the high cache loading overhead, InstGenIE employs a bubble-free pipeline scheme that overlaps computation with cache loading. Additionally, to reduce queuing latency in online serving while improving the GPU utilization, InstGenIE proposes a novel continuous batching strategy for diffusion model serving, allowing newly arrived requests to join the running batch in just one step of denoising computation, without waiting for the entire batch to complete. As heterogeneous masks induce imbalanced loads, InstGenIE also develops a load balancing strategy that takes into account the loads of both computation and cache loading. Collectively, InstGenIE outperforms state-of-the-art diffusion serving systems for image editing, achieving up to 3x higher throughput and reducing average request latency by up to 14.7x while ensuring image quality.

Dongqi Wang, Shengyu Zhang, Zhipeng Di et al.

Network pruning and knowledge distillation are two widely-known model compression methods that efficiently reduce computation cost and model size. A common problem in both pruning and distillation is to determine compressed architecture, i.e., the exact number of filters per layer and layer configuration, in order to preserve most of the original model capacity. In spite of the great advances in existing works, the determination of an excellent architecture still requires human interference or tremendous experimentations. In this paper, we propose an architecture slimming method that automates the layer configuration process. We start from the perspective that the capacity of the over-parameterized model can be largely preserved by finding the minimum number of filters preserving the maximum parameter variance per layer, resulting in a thin architecture. We formulate the determination of compressed architecture as a one-step orthogonal linear transformation, and integrate principle component analysis (PCA), where the variances of filters in the first several projections are maximized. We demonstrate the rationality of our analysis and the effectiveness of the proposed method through extensive experiments. In particular, we show that under the same overall compression rate, the compressed architecture determined by our method shows significant performance gain over baselines after pruning and distillation. Surprisingly, we find that the resulting layer-wise compression rates correspond to the layer sensitivities found by existing works through tremendous experimentations.