genai-workshop
This workshop is aimed to familarise users with the foundations of deploying and working with LLMs on Amazon SageMaker.
There are three labs in this workshop
Lab 1:
Contains 4 sections:
Section 1: a guide to show you how simply deploy LLMs from SM jumpstart UI
Section 2: demonstrates how to user Hugging face Deep Learning Container (DLC) to deploy modles on SageMaker. This section also includes promot templates and how to invoke the model for users who go down this section and deploy their own models
Section 3: demonstrates zero shot prompt with Flan_T5_XXL model to extract the information user need (in this scenario the intent or topic of a coversation)- Note, the Flan_T5_XXL model is deployed to a private account accessible via restful Api URL- the user will need a authentication pass code (not availble to public).
Section 4: demonstrates additional promot engineering via zero shot learning that allows the user to pass on specific labels and ask the model to pick one of those classes ( useful for hierarchy of intents/topics or sub intent/topics for example)- Note, the Flan_T5_XXL model is deployed to a private account accessible via restful Api URL- the user will need a authentication pass code (not availble to public).
Lab 2:
Walks you through some fundamental apis of langchain and how to usetilse them with SageMaker Models. Note, this workshop interacts with Falcon 40B Instruct model which is deployed to a private account accessible via restful Api URL- the user will need a authentication pass code (not availble to public).
Lab3:
Uses Lab 1 and Lab2 as a foundation to build and end to end RAG (Retreival Augmented Generation) application using Langchain and an LLM. Note, this workshop interacts with Falcon 40B Instruct model which is deployed to a private account accessible via restful Api URL- the user will need a authentication pass code (not availble to public).