Research map / Image generation / Fast diffusion sampling
Accelerating Diffusion via Hybrid Data-Pipeline Parallelism Based on Conditional Guidance Scheduling
Euisoo Jung, Byunghyun Kim, Hyunjin Kim, Seonghye Cho, Jae-Gil Lee
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adaptive parallelism switching on top of full condition-based partitioning (ablation) | Hybrid parallelism · SDXL 1024x1024 | speed-up: 1.78x -> 2.31x | Improved |
| add adaptive parallelism switching on top of full condition-based partitioning (ablation) | Hybrid parallelism · SDXL 1024x1024 | latency (s): 9.24 -> 7.12 | Improved |
| add adaptive parallelism switching on top of full condition-based partitioning (ablation) | Hybrid parallelism · SDXL 1024x1024 | FID w/ Orig.: 3.623 -> 4.100 | Hurt |
| increase parallelism interval k from 5 to 10 (ablation) | Hybrid parallelism · SDXL | FID w/ Orig.: 4.100 -> 5.942 | Hurt |
| increase parallelism interval k from 5 to 30 (ablation) | Hybrid parallelism · SDXL | speed-up: 2.31x -> 2.78x | Improved |
| increase parallelism interval k from 5 to 30 (ablation) | Hybrid parallelism · SDXL | FID w/ Orig.: 4.100 -> 9.191 | Hurt |
| increase parallelism interval k from 5 to 20 (ablation) | Hybrid parallelism · SDXL | latency (s): 7.12 -> 6.44 | Improved |
| use condition-based instead of patch-based partitioning with adaptive switching | Hybrid parallelism vs DistriFusion · SDXL | speed-up: 1.22x -> 2.31x | Improved |
| use hybrid parallelism instead of AsyncDiff pipeline | Hybrid parallelism vs AsyncDiff · SDXL | communication (GB): 9.830 -> 0.516 | Improved |
| use hybrid parallelism instead of AsyncDiff pipeline | Hybrid parallelism vs AsyncDiff · SD3 | speed-up: 1.97x -> 2.07x | Improved |
| use hybrid parallelism instead of Parastep | Hybrid parallelism vs Parastep · SD3 | FID w/ Orig.: 3.350 -> 1.878 | Improved |
| use hybrid parallelism instead of Parastep | Hybrid parallelism vs Parastep · SD3 | communication (GB): 0.032 -> 0.189 | Hurt |
Closest papers by meaning
- Communication-Efficient Diffusion Denoising Parallelization via Reuse-then-Predict Mechanism (NeurIPS 2025 · similarity 0.92)
- AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising (NeurIPS 2024 · similarity 0.92)
- DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Models (CVPR 2024 · similarity 0.90)
- PipeFusion: Patch-level Pipeline Parallelism for Diffusion Transformers Inference (NeurIPS 2025 · similarity 0.90)
- Accelerating Diffusion Models with Parallel Sampling: Inference at Sub-Linear Time Complexity (NeurIPS 2024 · similarity 0.90)
- Accelerating Parallel Diffusion Model Serving with Residual Compression (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
