Oracle Cloud Infrastructure Documentation

Fine-Tuning the Base Models in Generative AI

Create custom models in OCI Generative AI by fine-tuning the base models with your own dataset. You can create new custom models or create new versions of existing custom models.

Important

If you want to use a model that was already fine-tuned outside the service, you can import a compatible model from Hugging Face. OCI Generative AI doesn't fine-tune imported models. See Compatible Models for Import for the supported model list.

The following workflow displays the tasks for fine-tuning a pretrained model and hosting the fine-tuned model.

User workflow options for using the Generative AI service for fine-tuning a pretrained model and hosting the fine-tuned model through an endpoint.

The following tasks provide the steps for fine-tuning a pretrained model and hosting the fine-tuned model:

  1. Create the training dataset.
  2. Add the dataset to an Object Storage bucket.
  3. Create a fine-tuning dedicated AI cluster.
  4. Create a new custom model or create a new version from an existing model on the fine-tuning dedicated AI cluster.
  5. Create a hosting dedicated AI cluster.
  6. Create an endpoint for the custom model on the hosting dedicated AI cluster.
  7. Run the custom model in the playground.