Finetune

In Murnitur, fine-tuning prompts is a way to customize and improve the performance of your AI models based on specific needs. You can fine-tune prompts using preset configurations or by applying reinforcement learning techniques with your completion datasets.

​

Fine-Tuning with Presets

Murnitur provides preset configurations that can be fine-tuned to better fit your requirements. For example, you can load a preset designed for Question and Answer tasks and then fine-tune for better output:

preset = murnitur.load_preset("QA")
prompts = preset.fine_tune()  # Fine-tunes the preset with the current active version

If you want to use a specific version of the preset, you can pass the version as an argument.

​

Fine-Tuning with Datasets

For more advanced customization, you can use completion datasets to fine-tune prompts. This approach is similar to reinforcement learning, where the model is trained based on specific data to improve its responses:

prompts = preset.fine_tune_with_dataset('DATASET-ID')  # Fine-tunes the preset with a dataset

Replace 'DATASET-ID' with the identifier of your dataset. You can also pass a specific prompt version if needed.

​

Refining Prompts

You can further refine raw prompts without loading a preset. Use the refine_prompt method to make these adjustments:

prompts = murnitur.refine_prompt([
    {"role": "user", "content": "Answer this question: {{question}}, reference these contexts: {{contexts}}"}
])