当前位置:首页 > 报告详情

OCI Kubernetes Engine 助力大规模 AI 创新 [PAN1753].pdf

上传人: Fl****zo 编号:970889 2025-11-08 62页 3.57MB

1、 OCI Kubernetes Engine Powers AI Innovation at ScalePAN1753Kay Singh,Mickey BoxellOKE Product ManagerOracle AI World 25 Agenda title2Copyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly Restricted1234Why Kubernetes for AI workloads?Customer presentations(Faire,Pipefy)O

2、CIs AI InfrastructureOKE ecosystem for enabling AI workloadsYour hostsCopyright 2025,Oracle and/or its affiliates|Confidential:Internal/Restricted/Highly RestrictedKay SinghProduct Manager,OKEMickey BoxellProduct Manager,OKE3OKE customersCopyright 2025,Oracle and/or its affiliates|Confidential:Inter

3、nal/Restricted/Highly RestrictedManoj KrishnanSenior Staff ML Engineer,FaireSobhan DaliryChief Product Officer,Pipefy4AI Models,Training,and InferencingAI models are programs that learn from data to generate outputModel training is the process of teaching a model to recognize patterns and make predi

4、ctions by feeding it data and adjusting its parameters until it produces the desired behaviorTrained models are used to make inferences:generating output on novel input data based on its trainingExamples of models include Large Language Models(LLMs),Image and video recognition,Anomaly detection,Reco

5、mmender systems,Predictive modeling and forecasting,Robotics and control systemsKubernetes is the Platform for AI/ML WorkloadsTrain large-scale AI/ML models faster and more economicallyReduce the cost of training neural networksInfuse AI into apps using Cohere large language modelsPowering speech re

6、cognition for Mercedes and PandoraEmbed LLMs,narratives,and non-playable characters that evolve with each action for interactive gaming experiencesAI platform to collaborate and share,fine-tune,and run large language modelsOnline wholesale marketplace connecting independent brands with retail stores

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。
根据报告的内容,全文主要内容概括如下: - **Kubernetes在AI/ML工作负载中的重要性**:Kubernetes提供可扩展性、资源管理、生态系统和社区支持,是AI/ML工作负载的理想平台。 - **OCI Kubernetes Engine (OKE)的优势**:OKE提供灵活的基础设施、GPU优化、安全合规性、企业级支持,并支持Nvidia和AMD GPU。 - **案例研究**: - **Faire**:通过OKE优化GPU集群,提高搜索相关性预测准确性,并减少运行时间。 - **Pipefy**:利用OKE构建端到端信用工作流程,实现成本和效率提升。 - **OKE的AI工作负载支持**: - **计算**:支持多种GPU选项和裸金属形状。 - **网络**:提供高性能的RDMA网络,支持大规模集群。 - **存储**:提供高性能文件存储和对象存储,满足AI工作负载的存储需求。 - **软件**:支持Ubuntu工作节点镜像和GPU设备插件。
"OKE助力AI,效率翻倍?" "AI训练,OKE如何优化?" "GPU集群,OKE如何管理?"
客服
商务合作
小程序
服务号
折叠