SnapGen++: Unleashing Diffusion Transformers for Efficient High-Fidelity Image Generation on Edge Devices
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910019231940608 |
|---|---|
| author | Hu, Dongting Gupta, Aarush Gabidolla, Magzhan Sahni, Arpit Coskun, Huseyin Li, Yanyu Idelbayev, Yerlan Mahmood, Ahsan Lebedev, Aleksei Lahiri, Dishani Goyal, Anujraaj Hu, Ju Gong, Mingming Tulyakov, Sergey Kag, Anil |
| author_facet | Hu, Dongting Gupta, Aarush Gabidolla, Magzhan Sahni, Arpit Coskun, Huseyin Li, Yanyu Idelbayev, Yerlan Mahmood, Ahsan Lebedev, Aleksei Lahiri, Dishani Goyal, Anujraaj Hu, Ju Gong, Mingming Tulyakov, Sergey Kag, Anil |
| contents | Recent advances in diffusion transformers (DiTs) have set new standards in image generation, yet remain impractical for on-device deployment due to their high computational and memory costs. In this work, we present an efficient DiT framework tailored for mobile and edge devices that achieves transformer-level generation quality under strict resource constraints. Our design combines three key components. First, we propose a compact DiT architecture with an adaptive global-local sparse attention mechanism that balances global context modeling and local detail preservation. Second, we propose an elastic training framework that jointly optimizes sub-DiTs of varying capacities within a unified supernetwork, allowing a single model to dynamically adjust for efficient inference across different hardware. Finally, we develop Knowledge-Guided Distribution Matching Distillation, a step-distillation pipeline that integrates the DMD objective with knowledge transfer from few-step teacher models, producing high-fidelity and low-latency generation (e.g., 4-step) suitable for real-time on-device use. Together, these contributions enable scalable, efficient, and high-quality diffusion models for deployment on diverse hardware. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_08303 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SnapGen++: Unleashing Diffusion Transformers for Efficient High-Fidelity Image Generation on Edge Devices Hu, Dongting Gupta, Aarush Gabidolla, Magzhan Sahni, Arpit Coskun, Huseyin Li, Yanyu Idelbayev, Yerlan Mahmood, Ahsan Lebedev, Aleksei Lahiri, Dishani Goyal, Anujraaj Hu, Ju Gong, Mingming Tulyakov, Sergey Kag, Anil Computer Vision and Pattern Recognition Recent advances in diffusion transformers (DiTs) have set new standards in image generation, yet remain impractical for on-device deployment due to their high computational and memory costs. In this work, we present an efficient DiT framework tailored for mobile and edge devices that achieves transformer-level generation quality under strict resource constraints. Our design combines three key components. First, we propose a compact DiT architecture with an adaptive global-local sparse attention mechanism that balances global context modeling and local detail preservation. Second, we propose an elastic training framework that jointly optimizes sub-DiTs of varying capacities within a unified supernetwork, allowing a single model to dynamically adjust for efficient inference across different hardware. Finally, we develop Knowledge-Guided Distribution Matching Distillation, a step-distillation pipeline that integrates the DMD objective with knowledge transfer from few-step teacher models, producing high-fidelity and low-latency generation (e.g., 4-step) suitable for real-time on-device use. Together, these contributions enable scalable, efficient, and high-quality diffusion models for deployment on diverse hardware. |
| title | SnapGen++: Unleashing Diffusion Transformers for Efficient High-Fidelity Image Generation on Edge Devices |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.08303 |