5b6b901c80e87770dc836f0e64e6eeb3c073efef,census/keras/trainer/model.py,,to_savedmodel,#Any#Any#,85

Before Change

        sess=sess,
        tags=[tag_constants.SERVING],
        signature_def_map={
            signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY: signature}
    )
    builder.save()

After Change

def to_savedmodel(model, export_path):
  Convert the Keras HDF5 model into TensorFlow SavedModel.
  with K.get_session() as sess:
    saved_model_util.simple_save(
        sess,
        export_path,
        inputs={"input": model.inputs[0]},
        outputs={"income": model.outputs[0]})


def to_numeric_features(features):
  Convert the pandas input features to numeric values.

Instances