Inference Model


Inference is a package in Analytics Zoo aiming to provide high level APIs to speed-up development. It allows user to conveniently use pre-trained models from Analytics Zoo, Tensorflow and Caffe. Inference provides multiple Java/Scala interfaces.


  1. Easy-to-use java/scala API for loading and prediction with deep learning models.

  2. Support transformation of various input data type, thus supporting future prediction tasks

Primary APIs for Java


AbstractInferenceModel provides load API for loading a pre-trained model, thus we can conveniently load various kinds of pre-trained models in java applications. The load result of AbstractInferenceModel is a FloatInferenceModel. We just need to specify the model path and optionally weight path if exists where we previously saved the model.


AbstractInferenceModel provides predict API for prediction with loaded model. The predict result ofAbstractInferenceModel is a List<List<JTensor>> by default.

Java example

It's very easy to apply abstract inference model for inference with below code piece. You will need to write a subclass that extends AbstractinferenceModel.


public class TextClassificationModel extends AbstractInferenceModel {
    public TextClassificationModel() {
TextClassificationModel model = new TextClassificationModel();
model.load(modelPath, weightPath);
List<List<JTensor>> result = model.predict(inputList);

Primary APIs for Scala


InferenceSupportive is a trait containing several methods for type transformation, which transfer a model input to a valid data type, thus supporting future inference model prediction tasks.

For example, method transferTensorToJTensor convert a model input of data type Tensor to JTensor , which will be the input for a FloatInferenceModel.


FloatInferenceModel is an extending class of InferenceSupportive and additionally provides predict API for prediction tasks.


InferenceModelFactory is an object with APIs for loading pre-trained Analytics Zoo models, Caffe models and Tensorflow models. We just need to specify the model path and optionally weight path if exists where we previously saved the model. The load result of is a FloatInferenceModel.


ModelLoader is an extending object of InferenceSupportive and focus on the implementation of loading pre-trained models