当前位置:首页 > 报告详情

高通:2023年生成式人工智能在边缘(英文版)(37页).pdf

上传人: AG 编号:605584 2023-11-08 37页 8.03MB

1、Snapdragon and Qualcomm branded products are products of Qualcomm Technologies,Inc.and/or its subsidiaries.November 8,2023QCOMResearchQCOMResearchSnapdragon and Qualcomm branded products are products of Qualcomm Technologies,Inc.and/or its subsidiaries.Generative AI at the edgeJoseph Joseph SoriagaS

2、oriagaSenior Director,TechnologyQualcomm Technologies,Inc.2TodaysagendaWhy on-devicegenerative AI is keyFull-stack AI optimizationsfor diffusion models Stable Diffusion Stable Diffusion Full-stack AI optimizationsfor large language models Llama 2Llama 2Hybrid AI technologiesand architecturesQ&A23AIM

3、ET is a product of Qualcomm Innovation Center,Inc.Qualcomm AI Research is an initiative of Qualcomm Technologies,Inc.LLM:Large language mode;LVM:Language vision modelLeading machinelearning researchfor on-device AIacross the entirespectrum of topics Platform Platform researchresearchAppliedAppliedre

4、searchresearchFundamentalFundamentalresearchresearchAI researchAI researchGenerative AIG-CNNSelf-supervised learningReinforcement learningCausality and system-2Model quantization,compression,and NASHW-SW co-designDistillation of generative modelsPower managementAI Model EfficiencyToolkit(AIMET)Deep

5、learningfor 3D/geometryAudio and videocompressionAI for wirelessand RF sensingEnergy-efficientperceptionAI forchip designOn-device learningBayesian distributed learningGraph and kerneloptimizationFederated learningDeep learning for graphicsVideo recognition and predictionVirtual AI assistant(LLM)Dif

6、fusion-based image generation(LVM)Voice UI4FullFull-stack stack AI research&AI research&optimizationoptimizationModel,hardware,and softwareinnovation across each layerto accelerate AI applicationsEarly R&D andtechnology inventionsessential to leadingthe ecosystem forwardTransfer tech to commercialte

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
本文主要介绍了高通在边缘设备上实现生成式人工智能(Generative AI)的研究进展。主要内容包括: 1. 为什么在设备上实现生成式AI至关重要,以及高通在AI研究、应用研究和基础研究方面的全面布局。 2. 介绍了高通对Stable Diffusion和Llama 2等模型的全栈AI优化,包括模型压缩、量化、编译优化等,显著提升了在边缘设备上的推理速度和能效。 3. 展示了高通在边缘设备上实现的一些生成式AI应用,如文本到图像、语音助手等,以及高通在混合AI技术方面的探索。 4. 介绍了高通在AI研究方面的愿景,包括问题识别、生态系统合作、模型优化、发明创新等阶段,以及高通AI研究如何推动整个生态系统的发展。 5. 最后,高通表示其在边缘设备上实现生成式AI的领导地位,将有助于生成式AI的规模化应用,并展望了混合AI的未来。
如何在边缘设备上实现生成式AI? 混合AI技术如何推动生成式AI的发展? 如何在保持准确性的同时,减少生成式AI模型的内存占用?
客服
商务合作
小程序
服务号
折叠