Skip to content

Latest commit

 

History

History
71 lines (59 loc) · 2.34 KB

m2_studio_lang.md

File metadata and controls

71 lines (59 loc) · 2.34 KB

Module 2, Generative AI Studio Language Demo

Setup

  1. None

Instructions

Section 1

The goal in this section is to show how chat uses the context of the converation to shape answers. Context is provided via the context input, but also the history of the conversation.

  1. Navigate to Vertex AI > Generative AI Studio > Language > Start a conversation > TEXT CHAT
  2. Enter the following into the Context area: 
    Your name is Miles. You are an astronomer who is knowledgeable about the solar system.
Respond in short sentences. Shape your response as if talking to a 10-years-old.
  3. Start the conversation by asking: 
    How many planets are there in the solar system?
  4. Follow up by asking several follow up questions: 
    What are their names?

    Which is the biggest?
  5. Point out that when you ask "what are their name?" it knows you're talking about the eight planets in the solar system.
  6. Point out that you can save your chat. Ask why you might want to do that?

Section 2

The point of this section is to show how changing the temperature changes the consistency and creativity of the responses provided.

  1. Clear the conversation

  2. Set the following parameters (leave others at default values):

    Parameter Value
    Temperature 0
    Token Limit 1024

  3. Start the conversation with the following prompt:

    Tell me about the solar system

  4. Ask the same question a few more times. Point out that the answer is identical each time. Tie that back to temperature.

  5. Clear the conversation

  6. Set the following parameters (leave others at default values):

    Parameter Value
    Temperature 1

  7. Start the conversation with the following prompt:

    Tell me about the solar system

  8. Ask the same question a few more times. Point out that the answer is different each time and note the differences. Tie that back to temperature.

Section 3

The point of this section is to show how both the prompt gallery and how you can use the structured prompt design UI to generate content.

  1. Walk the students quickly through https://console.cloud.google.com/vertex-ai/generative/language/prompt-examples/Hashtag%20tokenization