Training on Logs

In FinetuneDB, you can search and filter past LLM requests to inspect responses and build a training dataset.

Identify Logs

Logs are systematically tracked for future fine-tuning, capturing feedback, tags, and anomalies. Use the Log Viewer to identify relevant logs. Apply filters such as model and custom tags to locate the data needed for fine-tuning.

Add Logs to Dataset

Select the filtered logs and import them into a dataset. Save the dataset to use it for fine-tuning your models.

Train on Logs

By leveraging production data, you can speed up the creation of fine-tuning datasets. This approach also facilitates switching between models, such as moving from a proprietary model (like GPT-4) to an open-source model.