当前位置:首页 > 报告详情

使用 NVIDIA Lepton on OCI 释放您的 GPU 性能 [LRN3343].pdf

上传人: Fl****zo 编号:970988 2025-11-08 13页 576.81KB

1、Oracle AI World 2025Jake Bloom(NVIDIA)DGX Cloud Product ManagementTaylor Newill(Oracle)OCI Strategic InitiativesAI Developers ChallengesScaling workloads across multiple regions and clouds is complexChallenging to discover GPUs based on region,cost,and performanceAdministration across multiple Cloud

2、s is hardFragmented Experience for developing,customizing and deploying appsDGX Cloud LeptonConnecting Developers to GPU Compute An AI platform that connects developers with GPU compute across a network of cloud providers Access Global GPU ComputePlatform for Developing,Customizing&Deploying Applica

3、tionsNVIDIA CONFIDENTIAL.DO NOT DISTRIBUTE.InferenceDeveloper ToolsTrainingDesigned for Developers Unified experience to easily access all services Train and Scale Across Clouds With The DGX Cloud Lepton AI PlatformDevelopment Run interactive development sessions,SSH,Jupyter notebooks,VS CodeInferen

4、ce and Endpoints Fast and scalable inference across multiple clusters and regions powered by NVIDIA Cloud Functions(NVCF).Easily create NVIDIA NIM endpoints Training and Fine Tuning Run distributed training or batch processing jobs,with high performance interconnects and accelerated storageGPU-Bare

5、Metal or VM InstanceNetworkingNVIDIA Cloud Partners StorageNVIDIA AI EnterpriseAI&Data Science Development&Deployment ToolsDGX Cloud Lepton AI StackBatch JobsAI Infrastructure ManagementGPU Cloud ProvidersDev PodsEndpointsHealth MonitoringObservabilityResilience Compute Resource Management The DGX C

6、loud Lepton stackEnterprise grade stack to support your AI workloadsBYO ComputeNvidia Managed CapacityOn-Demand*Monitor SystemMonitor overall system metrics(CPU,memory,disk).Monitor MetricsMonitor critical GPU and GPU fabric metrics(power,temperature)Report StatusReports GPU and GPU fabric status(nv

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《Data》标记内容,全文主要内容概括如下: - **DGX Cloud Lepton平台**:一个连接开发者与GPU计算的AI平台,支持跨云开发、定制和部署应用。 - **关键点**: - **跨云工作负载**:简化跨多个区域和云的负载扩展。 - **GPU访问**:根据区域、成本和性能发现GPU。 - **统一体验**:提供统一的开发、定制和部署体验。 - **开发工具**:支持SSH、Jupyter笔记本、VS Code等开发工具。 - **推理和训练**:提供快速和可扩展的推理和分布式训练。 - **监控和可观察性**:实时监控作业,提供健康检查和错误检测。 - **企业级堆栈**:支持企业级AI工作负载。 - **GPU管理**:使用GPUd工具减少GPU集群不可用性。 - **客户工作流程**:提供早期访问的客户工作流程,包括数据训练、推理和GPU分配。 - **外部存储选项**:支持从不同位置引入数据,包括对象存储和文件存储。
AI开发新体验?" "跨云GPU计算,挑战与解决方案?" "NVIDIA AI平台,开发者必备工具?"
客服
商务合作
小程序
服务号
折叠