1465 - 利用 IBM 云和可部署架构上的 Intel Gaudi 3 加速您的 AI 部署.pdf

1、Fast Track your AI Deployment with Intel Gaudi 3 Fast Track your AI Deployment with Intel Gaudi 3 Deployable Architectures on IBM CloudDeployable Architectures on IBM CloudSession#1465IBM TechXchange|2025 IBM Corporation1#IBMTechXchangeRyan ParkRyan ParkField Applications EngineerIntel CorporationPr

2、amod PaiPramod PaiSoftware EngineerIntel CorporationBrian SathianathanBrian SathianathanFounder and CTOIterateAIWhat you will learn in this session0102030405What is Intel Gaudi 3?What is Intel AI for Enterprise Inference?Easily deploy with Deployable ArchitecturesDemo:Enterprise Inference in ActionA

3、gentOne by IterateAIIBM TechXchange|2025 IBM CorporationIntel Gaudi 3 AI Accelerators on IBM CloudBringing choice of generative AI inferencing to mainstream enterprisesIBM TechXchange|2025 IBM Corporation3Competitive generativeAI performanceCost-conscious entry pointOpen developmentEfficient and sec

4、ure scalingIBM Cloud is the first global cloud provider to widely adopt Intel Gaudi 3 AI accelerators1 IBM Cloud Virtual Servers for VPCRed Hat OpenShift on IBM CloudIBMwatsonxDeployable ArchitecturesIBM TechXchange|2025 IBM Corporation4 500K+Transformer and Diffusion models on Hugging Face are easi

5、ly enabled on the Intel Gaudi platform Composable Microservices with OPEA Enterprise RAG&Inference API End Point use case solutions Learn moreIntel Gaudi 3 AI Accelerator Performance DataHugging Face Optimum for Intel Gaudi 3 AI AcceleratorsIntel Gaudi Accelerator Validated Hugging Face ModelsIntel

6、Gaudi Accelerator Model CatalogIntel Gaudi Software Support MatrixRepresentative Models on Intel Gaudi PlatformLlaMA 2,3,4Stable DiffusionMPTWave2vec2DeepSeek distilledMistralPhi-3WhisperLLaVaMixtralCode LlamaQwen2GraniteFalconStarCoderBridgetowerFrameworks&Libraries OrchestrationToolsMegatron-LMTra