How to load a saved model which is saved from python and predict in new tensorflow java ?

[DachuanZhao]

System information

• TensorFlow version: 2.5.0
• Doc Link: https://www.tensorflow.org/jvm/install

Describe the documentation issue The java document only contains a install chapter , could you please add a load saved model and predict chapter ?

[Craigacp]

You can see a short example of loading in a saved model and executing a function from it in the TF-Java test suite (https://github.com/tensorflow/java/blob/master/tensorflow-core/tensorflow-core-api/src/test/java/org/tensorflow/SavedModelBundleTest.java#L285), and there are many more examples of that in the TF-Java models repo (https://github.com/tensorflow/java-models/tree/master/tensorflow-examples/src/main/java/org/tensorflow/model/examples).

We do need to improve the public facing docs, as admittedly these things are not easy to find if you start on tensorflow.org.

[luobotaxinghu]

You can see a short example of loading in a saved model and executing a function from it in the TF-Java test suite (https://github.com/tensorflow/java/blob/master/tensorflow-core/tensorflow-core-api/src/test/java/org/tensorflow/SavedModelBundleTest.java#L285), and there are many more examples of that in the TF-Java models repo (https://github.com/tensorflow/java-models/tree/master/tensorflow-examples/src/main/java/org/tensorflow/model/examples).

We do need to improve the public facing docs, as admittedly these things are not easy to find if you start on tensorflow.org.

This example TF-Java test suite load from a dir, have some example that load byte[] of .pb to SavedModelBundle?

[Craigacp]

You can't do that as the SavedModelBundle format is a pb along with a checkpoint of the variables (well, it's conceptually a checkpoint). If you've got a frozen graph from TF 1 then you can load that directly as a GraphDef and import it into a Graph.