Selecting a foundation model

Introduction to Amazon Bedrock

Nikhil Rangarajan

Data Scientist

Enabling model access in Amazon Bedrock

Access model settings - no need to enable models for the course

A screenshot showing the Claude and Nova models enabled on the AWS Console.

Lightweight vs. advanced foundation models

Lightweight models
- Best for: Q&A, summarization
- Quick response times and cost-effective
- Example: Nova Micro, Claude Haiku

Logo of Amazon.

Advanced models
- Best for: Complex reasoning, analysis
- Comprehensive analysis capabilities
- Example: Nova Plus, Claude Sonnet

Other foundation models in Amazon Bedrock

Other models available in Amazon Bedrock:

Meta's Llama
- 📝 Open-source LLM for general text tasks
Stability AI's Stable Diffusion
- 🖼 Specialized for image generation
AI21's Jurassic
- 📈 Text generation and analysis

Meta's llama logo.

Invoking Bedrock models

Basic structure:

import json

bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')

response = bedrock.invoke_model(modelId='amazon.nova-lite-v1:0',

body=json.dumps(input_dictionary)

)

Returns model's response in JSON format

Invoking Claude

Anthropic Claude

response = bedrock.invoke_model(
    modelId='anthropic.claude-3-5-sonnet-v2:0',
    body=json.dumps(

      {"anthropic_version": "bedrock-2023-05-31",

       "max_tokens": 100,

       "messages": [{
            "role": "user",
            "content": [{"type": "text", "text": "your prompt here"}],
        }],

    }

))

Invoking Nova

Amazon Nova

response = bedrock.invoke_model(
    modelId='amazon.nova-lite-v1:0', 
    body=json.dumps(

      {"messages":

                  [{"role": "user", 
                    "content": [{"text": "your prompt here"}]
                   }]

      }

))

Extracting model response

Read the body of the API response
Convert from JSON to a Python dictionary
Extract output using keys

print("Claude:",

      json.loads(
          claude_response['body'].read()
        )

      ["content"][0]["text"]

     )

Hello! It's nice to meet you. How can I assist you today?

Extracting model response

Read the body of the API response
Convert from JSON to a Python dictionary
Extract output using keys

print("Nova:",

      json.loads(
        nova_response.get("body").read().decode()
      )

      ["output"]["message"]["content"][0]["text"]

     )

Hello! How can I assist you today?

Let's practice!

Introduction to Amazon Bedrock