Fine-tuning workflow

The fine-tuning process is comprised of the following steps:

1. Prepare and upload training data: You prepare the taxonomies the model must learn and securely share the training dataset using private links.
2. Train the new fine-tuned model: TwelveLabs’s automated pipeline trains the model using your dataset.
3. Evaluate the fine-tuned model: You evaluate the fine-tuned model quantitatively and qualitatively.
   1. Quantitative evaluation: Assess the model’s performance using metrics such as mean average precision (mAP). TwelveLabs will provide the metrics for the private beta release.
      Qualitative evaluation: The fine-tuned model will be available in the Playground or accessible programmatically through the API for qualitative assessment.
4. Deploy the fine-tuned model: TwelveLabs deploys the fine-tuned model to your environment.
   Based on the evaluation results, you can retrain the model by repeating the cycle from the second step to improve its performance.