Create a Model Configuration

After you add a foundation model, you can create an associated model configuration for a specific use case.

1. In AI Models (formerly Einstein Studio), click New Configuration.
2. In the Model Playground, select Create model.
3. Select the foundation model that you want to use.
4. In Advanced Settings, define the hyperparameters that control your model’s learning process.
   These values are important for training and influencing prompt responses.

   • Temperature

     Determines the creativity of a response with a value from 0 through 2. A lower value generates more consistent responses. A higher value (such as 0.7-0.9) generates more random responses. Specify a value based on how coherent or creative you want the generated text to be.

     Only Anthropic models have this setting available for testing.

   • Frequency Penalty

     Discourages frequent use of the same tokens (words, phrases, or code) in a response. Specify a value from -2.0 through 2.0. To reduce repetition in the generated text, use a higher value.

   • Presence Penalty

     Encourages the use of different tokens in a response. Specify a value from -2.0 through 2.0. To include tokens that haven’t been used in the generated text, use a higher value.

5. Define your prompt settings.
   • Go to the Prompt section, and click Settings.
   • Specify the stop sequence and data masking for each prompt that you enter.
6. To start over and define your model configuration settings, click Restore Default.
7. To exit Model Playground and view your model on the Generative tab, click Close.