Fine-tune overview

You can fine-tune GPT models through OpenAI, and Baserun also offers services for fine-tuning open source models through our partner, OpenPipe. Please read our blog post to learn more about when to fine-tune open source models.

​

Curating Your Dataset from Logs:

Navigate to the Monitoring > LLM Request tab or Traces tab.

Filter the LLM requests or traces by the trace name or add filters. You can also use the evaluation feature and appropriate labels as identifiers to curate the relevant datasets.

Then click the export button. You will see the ‘Export Training Data’ option. This option will generate a JSON-formatted data table that can be uploaded into OpenAI or our fine-tuning partner, OpenPipe.

​

Fine-Tuning Models:

• OpenAI fine-tuning guide

• OpenPipe upload data, fine-tuning guide

​

Monitoring Custom Models:

Once you have finished fine-tuning, you can use OpenAI or OpenPipe’s endpoint to call your custom model. No changes are needed to monitor your custom models in production.

​

Use Custom Models in the Baserun Playground:

You must add the custom model in the settings to use your fine-tuned model in the Baserun playground. Follow the ‘Add Custom Model’ guide to add your fine-tuned model.