This page show you how to find and use a trained machine learning (ML) model on AI Hub.
Find and use a trained model
To find a trained model:
- Open your web browser and go to AI Hub.
- Browse or search for assets, then click Trained model under Category to filter the list of assets to display only trained models.
- Once you find the model you are interested in, click the model's name. AI Hub displays a description of the model and information on how to use it. For more information on how to find assets, read the guide to finding assets on AI Hub.
Click Download under Use this asset. A zip file containing the model downloads to your computer.
- TensorFlow models are saved as a language-neutral serialized representation of the model's variables, graph, and the graph's metadata. For more information, see the TensorFlow guide to the SavedModel format.
XGBoost and scikit-learn models are persisted by serializing the trained model object. Models that are saved with the pickle module are saved as pkl files. Models that are saved with the joblib module are saved as joblib files. XGBoost models can also be persisted with the xgboost.Booster class as bst files.
You can now unzip the model and experiment with it on your computer, or you can deploy the model as an ML service for predictions. To learn more, read the AI Platform guide to deploying a model to serve predictions.
Before you experiment with the model, read the model description in AI Hub to understand the following:
- The model's inputs and outputs.
- Any preprocessing steps that the model depends on.
- The environment that the model expects to run in, including the version of Python, the ML framework, and the version of the ML framework.