LIVIVO - The Search Portal for Life Sciences

zur deutschen Oberfläche wechseln
Advanced search

Search results

Result 1 - 3 of total 3

Search options

  1. Book ; Online: Infinite-LLM

    Lin, Bin / Peng, Tao / Zhang, Chen / Sun, Minmin / Li, Lanbo / Zhao, Hanyu / Xiao, Wencong / Xu, Qi / Qiu, Xiafei / Li, Shen / Ji, Zhigang / Li, Yong / Lin, Wei

    Efficient LLM Service for Long Context with DistAttention and Distributed KVCache

    2024  

    Abstract: The rapid proliferation of Large Language Models (LLMs) has been a driving force in the growth of cloud-based LLM services, which are now integral to advancing AI applications. However, the dynamic auto-regressive nature of LLM service, along with the ... ...

    Abstract The rapid proliferation of Large Language Models (LLMs) has been a driving force in the growth of cloud-based LLM services, which are now integral to advancing AI applications. However, the dynamic auto-regressive nature of LLM service, along with the need to support exceptionally long context lengths, demands the flexible allocation and release of substantial resources. This presents considerable challenges in designing cloud-based LLM service systems, where inefficient management can lead to performance degradation or resource wastage. In response to these challenges, this work introduces DistAttention, a novel distributed attention algorithm that segments the KV Cache into smaller, manageable units, enabling distributed processing and storage of the attention module. Based on that, we propose DistKV-LLM, a distributed LLM serving system that dynamically manages KV Cache and effectively orchestrates all accessible GPU and CPU memories spanning across the data center. This ensures a high-performance LLM service on the cloud, adaptable to a broad range of context lengths. Validated in a cloud environment with 32 NVIDIA A100 GPUs in configurations from 2 to 32 instances, our system exhibited 1.03-2.4x end-to-end throughput improvements and supported context lengths 2-19x longer than current state-of-the-art LLM service systems, as evidenced by extensive testing across 18 datasets with context lengths up to 1,900K.
    Keywords Computer Science - Distributed ; Parallel ; and Cluster Computing ; Computer Science - Hardware Architecture
    Subject code 303
    Publishing date 2024-01-05
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  2. Article: Phosphorus/nitrogen co-doped and bimetallic MOF-derived cathode for all-solid-state rechargeable zinc–air batteries

    Yang, Xing / Wu, Xianghua / Guo, Zeping / Li, Qingyu / Wang, Hongqiang / Ke, Chujun / Zeng, Wei / Qiu, Xiafei / He, Yun / Liang, Xiaoguang / Kim, Yoonseob

    RSC advances. 2020 Sept. 10, v. 10, no. 55

    2020  

    Abstract: With the merits of high safety and energy density, all-solid-state zinc–air batteries possess potential applications in flexible and wearable electronic devices. Especially, the air cathodes with bifunctional catalytic activity, i.e. oxygen reduction ... ...

    Abstract With the merits of high safety and energy density, all-solid-state zinc–air batteries possess potential applications in flexible and wearable electronic devices. Especially, the air cathodes with bifunctional catalytic activity, i.e. oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) have been received enormous attention. In this work, we provide a novel phosphorus/nitrogen co-doped and bimetallic metal–organic framework (MOF)-derived cathode configurated with phosphorus-doped bimetallic FeNi alloys and a nitrogen-doped porous carbon layer loaded on graphene (P–FeNi/NC@G). The P–FeNi/NC@G electrode exhibits a superior OER activity with an overpotential of 310 mV at 10 mA cm⁻² and an ORR performance with a half-wave potential of 0.81 V. With P–FeNi/NC@G as the air cathode, the integrated all-solid-state rechargeable zinc–air battery presents a high open-circuit voltage of 1.53 V, a high peak power density of 159 mW cm⁻², a small charge–discharge voltage gap of 0.73 V at 5 mA cm⁻², as well as excellent long-term stability up to 144 cycles. This work not only expands the air cathode materials database but also develops a new co-doped synthesis method that can be utilized to fabricate a cathode with promoted catalytic efficiency, resulting in improved performance for an all-solid-state zinc–air battery.
    Keywords air ; alloys ; batteries ; catalytic activity ; cathodes ; coordination polymers ; databases ; density ; electric potential difference ; electrochemistry ; electronic equipment ; energy density ; exhibitions ; graphene ; nitrogen ; oxygen production ; phosphorus ; synthesis
    Language English
    Dates of publication 2020-0910
    Size p. 33327-33333.
    Publishing place The Royal Society of Chemistry
    Document type Article
    Note NAL-light
    ISSN 2046-2069
    DOI 10.1039/d0ra04827e
    Database NAL-Catalogue (AGRICOLA)

    More links

    Kategorien

  3. Article ; Online: Phosphorus/nitrogen co-doped and bimetallic MOF-derived cathode for all-solid-state rechargeable zinc-air batteries.

    Yang, Xing / Wu, Xianghua / Guo, Zeping / Li, Qingyu / Wang, Hongqiang / Ke, Chujun / Zeng, Wei / Qiu, Xiafei / He, Yun / Liang, Xiaoguang / Kim, Yoonseob

    RSC advances

    2020  Volume 10, Issue 55, Page(s) 33327–33333

    Abstract: With the merits of high safety and energy density, all-solid-state zinc-air batteries possess potential applications in flexible and wearable electronic devices. Especially, the air cathodes with bifunctional catalytic activity, ...

    Abstract With the merits of high safety and energy density, all-solid-state zinc-air batteries possess potential applications in flexible and wearable electronic devices. Especially, the air cathodes with bifunctional catalytic activity,
    Language English
    Publishing date 2020-09-10
    Publishing country England
    Document type Journal Article
    ISSN 2046-2069
    ISSN (online) 2046-2069
    DOI 10.1039/d0ra04827e
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

To top