Enhanced customization of API parameters for better control over the behavior of the Ollama model.

[moyueheng]

Features

• Enhanced customization of API parameters for better control over the behavior of the Ollama model.
• Added repeat_penalty parameter to the _llama_api_kwargs to reduce the likelihood of repetitive responses.

Feature Docs

• Detailed parameter usage and values can be found in the Ollama documentation: Ollama Model Documentation

Influence

• This change allows developers to fine-tune the model output based on specific needs by adjusting temperature and repeat_penalty dynamically through config.
• It should not affect existing implementations as long as the config object is properly updated to include repetition_penalty.

Result

• Unit tests were updated to cover new configuration options.
• All tests pass with the new changes, ensuring that existing functionalities remain unaffected.

Other

• Ensure that the config object in your application is updated to include repetition_penalty to avoid runtime errors.
• The implementation follows the recommended practices as per the latest Ollama API guidelines.