First Principles: Exploring the Depths of OCI Generative AI Service

Показать описание

This episode of First Principles covers the mechanics behind Oracle Cloud Infrastructure (OCI) Generative AI Service and how it makes it easier to work with large language models (LLMs) for AI applications. SVP and Chief Technical Architect for OCI, Pradeep VIncent, explains Oracle's upgrades in flexible fine-tuning of LLMs and improved GPU cluster management for simplified and cost-effective integration of LLMs to AI applications.

00:00 - Intro to OCI Generative AI Service
01:29 - What are Generative AI Models?
02:01 - Transformer Model Architecture
02:38 - Encoder-Decoder Transformer Model
03:35 - Gen AI in enterprise applications
04:05 - Gen AI key to success
04:05 - Achieving high accuracy of LLM outputs
06:08 - Retrieval Augmented Generations
06:30 - Basic OCI Gen AI Workflow
08:38 - Dedicated GPU RDMA clusters
09:34 - Customer Data Privacy and Security
10:45 - Fine-Tuning the Models
11:41 - Efficient Fine-Tuning w/ T-Few
12:10 - How T-Few Fine-Tuning Works
13:15 - Inside the Transformer Layer
14:42 - OCI Gen AI's cost effective inferencing
15:29 - Packing many models in single GPU cluster
17:45 - Key Takeaways

Follow us here:

#GenerativeAI #Oracle #OCI