Difference between revisions of "Resource:Seminar"

From MobiNetS
Jump to: navigation, search
 
(116 intermediate revisions by 4 users not shown)
Line 1: Line 1:
{{SemNote
{{SemNote
|time='''Friday 10:30-12:00'''
|time='''2026-04-10 10:30'''
|addr=4th Research Building A518
|addr=4th Research Building A518
|note=Useful links: [[Resource:Reading_List|Readling list]]; [[Resource:Seminar_schedules|Schedules]]; [[Resource:Previous_Seminars|Previous seminars]].
|note=Useful links: [[Resource:Reading_List|📚 Readling list]]; [[Resource:Seminar_schedules|📆 Schedules]]; [[Resource:Previous_Seminars|🧐 Previous seminars]].
}}
}}


===Latest===
===Latest===
{{Latest_seminar
{{Latest_seminar
|abstract=Continual learning (CL) trains NN models incrementally from a continuous stream of tasks. To remember previously learned knowledge, prior studies store old samples over a memory hierarchy and replay them when new tasks arrive. Edge devices that adopt CL to preserve data privacy are typically energy-sensitive and thus require high model accuracy while not compromising energy efficiency, i.e., cost-effectiveness. Our work is the first to explore the design space of hierarchical memory replay-based CL to gain insights into achieving cost-effectiveness on edge devices. We present Miro, a novel system runtime that carefully integrates our insights into the CL framework by enabling it to dynamically configure the CL system based on resource states for the best cost-effectiveness. To reach this goal, Miro also performs online profiling on parameters with clear accuracy-energy trade-offs and adapts to optimal values with low overhead. Extensive evaluations show that Miro significantly outperforms baseline systems we build for comparison, consistently achieving higher cost-effectiveness.
|abstract = To effectively utilize heterogeneous specialized hardware units in modern GPUs, such as TensorCores and Tensor Memory Accelerators, this paper introduces PipeThreader, a new DNN compiler. PipeThreader proposes shifting scheduling functionality from hardware to software so as to enable more efficient and sophisticated computation pipelining with minimal manual effort. This is achieved through sTask-graph, a new DNN computation abstraction, a hierarchical hardware abstraction that captures the capabilities of specialized units, and new scheduling primitives. As a result, PipeThreader can discover efficient pipeline scheduling for well-studied DNN architectures like FlashAttention, achieving comparable or even superior performance. Additionally, it can uncover novel pipeline schemes for emerging models like Mamba2, delivering significantly better performance compared to state-of-the-art hand-crafted implementations. The code is open-sourced at https://github.com/tile-ai/tilelang.
|confname=MobiCom'23
|confname =OSDI'25
|link=https://arxiv.org/pdf/2308.06053
|link = https://www.usenix.org/conference/osdi25/presentation/cheng
|title=Cost-effective On-device Continual Learning over Memory Hierarchy with Miro
|title= PipeThreader: Software-defined pipelining for efficient DNN execution
|speaker=Jiale
|speaker=Junzhe
|date=2024-06-14}}
|date=2026-4-9
{{Latest_seminar
}}
|abstract=Multi-view 3D reconstruction driven augmented, virtual, and mixed reality applications are becoming increasingly edge-native, due to factors such as, rapid reconstruction needs, security/privacy concerns, and lack of connectivity to cloud platforms. Managing edge-native 3D reconstruction, due to edge resource constraints and inherent dynamism of ‘in the wild’ 3D environments, involves striking a balance between conflicting objectives of achieving rapid reconstruction and satisfying minimum quality requirements. In this paper, we take a deeper dive into multi-view 3D reconstruction latency-quality trade-off, with an emphasis on reconstruction of dynamic 3D scenes. We propose data-level and task-level parallelization of 3D reconstruction pipelines, holistic edge system optimizations to reduce reconstruction latency, and long-term minimum reconstruction quality satisfaction. The proposed solutions are validated through collection of real-world 3D scenes with varying degree of dynamism that are used to perform experiments on hardware edge testbed. The results show that our solutions can achieve between 50% to 75% latency reduction without violating long term minimum quality requirements.
 
|confname=SEC'23
|link=https://www.cs.hunter.cuny.edu/~sdebroy/publication-files/SEC2023_CR.pdf
|title=On Balancing Latency and Quality of Edge-Native Multi-View 3D Reconstruction
|speaker=Yang Wang
|date=2024-06-14}}
{{Resource:Previous_Seminars}}
{{Resource:Previous_Seminars}}

Latest revision as of 10:37, 10 April 2026

Time: 2026-04-10 10:30
Address: 4th Research Building A518
Useful links: 📚 Readling list; 📆 Schedules; 🧐 Previous seminars.

Latest

  1. [OSDI'25] PipeThreader: Software-defined pipelining for efficient DNN execution, Junzhe
    Abstract: To effectively utilize heterogeneous specialized hardware units in modern GPUs, such as TensorCores and Tensor Memory Accelerators, this paper introduces PipeThreader, a new DNN compiler. PipeThreader proposes shifting scheduling functionality from hardware to software so as to enable more efficient and sophisticated computation pipelining with minimal manual effort. This is achieved through sTask-graph, a new DNN computation abstraction, a hierarchical hardware abstraction that captures the capabilities of specialized units, and new scheduling primitives. As a result, PipeThreader can discover efficient pipeline scheduling for well-studied DNN architectures like FlashAttention, achieving comparable or even superior performance. Additionally, it can uncover novel pipeline schemes for emerging models like Mamba2, delivering significantly better performance compared to state-of-the-art hand-crafted implementations. The code is open-sourced at https://github.com/tile-ai/tilelang.

History

2024

2023

2022

2021

2020

  • [Topic] [ The path planning algorithm for multiple mobile edge servers in EdgeGO], Rong Cong, 2020-11-18

2019

2018

2017

Instructions

请使用Latest_seminar和Hist_seminar模板更新本页信息.

    • 修改时间和地点信息
    • 将当前latest seminar部分的code复制到这个页面中
    • 将{{Latest_seminar... 修改为 {{Hist_seminar...，并增加对应的日期信息|date=
    • 填入latest seminar各字段信息
    • link请务必不要留空，如果没有link则填本页地址 https://mobinets.org/index.php?title=Resource:Seminar
  • 格式说明
    • Latest_seminar:

{{Latest_seminar
|confname=
|link=
|title=
|speaker=
}}

    • Hist_seminar

{{Hist_seminar
|confname=
|link=
|title=
|speaker=
|date=
}}

Retrieved from "http://mobinets.cn/site/index.php?title=Resource:Seminar&oldid=3511"