AI-901 ยท Generative AI apps in Foundry
Deploy a model and test it in the playground
Selecting a model from the catalog, deploying it, and using the chat playground to test prompts, adjust parameters and view code.
Key points
- Pick a model from the Foundry Models catalog and deploy it to your Foundry resource; you land in its playground.
- There you set a system prompt, adjust parameters such as temperature, top_p and max_tokens, compare up to three models side by side, and open the Code tab for ready-made samples before building the app.
Source: https://learn.microsoft.com/en-us/azure/foundry/concepts/concept-playgrounds
Practice questions
Where do you browse available models before choosing one to deploy in the Foundry portal?
- The Foundry Models catalog
- The Azure Marketplace only
- The Agents list of a project
- The Azure Policy blade
Show answer
Answer: A. The Foundry Models catalog
The Foundry Models catalog lists models from Microsoft, OpenAI, Meta and others. The agents list shows agents, not models to deploy.
Which settings can you tune while experimenting in the model playground?
- Temperature
- The number of GPUs backing the model
- Top_p
- Maximum tokens in the response
- The datacenter rack used for inference
Show answer
Answer: A. Temperature, C. Top_p, D. Maximum tokens in the response
The model playground exposes sampling parameters (temperature, top_p, max_tokens) and the system prompt. The Azure region of the model hardware and physical GPU count are not playground settings.
True or false: The playground's Code tab only offers sample code in Python.
Show answer
Answer: False
False. The Code tab provides multilingual samples (you pick the language) and can open them in VS Code for the Web.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
