With the MLOps capabilities provided by PolarDB for AI, you can upload, deploy, and use externally trained models.
Supported model types
A wide array of external models are supported. The following table lists examples of supported frameworks:
| Framework | Type |
|---|---|
| scikit-learn (sklearn)-compatible algorithms — LightGBM, Gradient Boosted Decision Tree (GBDT) | Traditional ML |
| TensorFlow | Deep learning |
| PyTorch | Deep learning |
Steps
Complete the following steps in order:
Upload a model — store the trained model file in PolarDB for AI.
Deploy a model — activate the uploaded model so it is ready to serve inference requests.
Evaluate a model — assess model performance against a test dataset before running it in production.
Perform model inference — run predictions against your data using SQL.