TensorFlow server

Model Information

Basic

Property	Description
Model Image	infuseai/tensorflow2-prepackaged:v0.2.0
Input	ndarray or image
Output	ndarray
Repository	Link

Compatibility of TensorFlow 2

Model Format	Support
SavedModel	Yes
HDF5	Yes

Compatibility of TensorFlow 1

Model Format	Support
*.pb	No
checkpoint	No
SavedModel	No
HDF5	Yes

Model URI Structure

SavedModel Format

We support TensorFlow2 SavedModel format. The model uri structure is just the output of tf.saved_model.save().

<model uri>
├── saved_model.pb
└── variables
    ├── variables.data-00000-of-00001
    └── variables.index

HDF5 Format

We also support HDF5 format which is saved from Keras API in both TensorFlow 2 and TensorFlow 1.

<model uri>
└── model.h5

1. model.h5: The file is HDF5 format, and can be any file name with .h5 file extension.

MLflow model

We also support MLflow model in Tensorflow Flavor and Keras Flavor which are exported from MLflow autologging API.

<model uri>
├── MLmodel
└── <model files>

How It Works

You can check the detailed code in the Github. Here, we brief the code as follows.

Load the model

def load(self):
    model_uri = self.model_uri
    # check model exported from mlflow.tensorflow.autolog()
    if os.path.isfile(os.path.join(model_uri, 'MLmodel')):
        if os.path.isdir(os.path.join(model_uri, 'data/model')):
            print("Loading model from tensorflow.keras.Model.fit + mlflow.tensorflow.autolog()")
            model_uri = os.path.join(model_uri, 'data/model')
        elif os.path.isdir(os.path.join(model_uri, 'tfmodel')):
            print("Loading model from tensorflow.estimator.Estimator.train + mlflow.tensorflow.autolog()")
            model_uri = os.path.join(model_uri, 'tfmodel')

    self.use_keras_api = 1
    if tf.saved_model.contains_saved_model(model_uri):
        self.model = tf.saved_model.load(model_uri).signatures["serving_default"]
        if 'saved_model' not in str(type(self.model)):
            self.use_keras_api = 0
        else:
            del self.model
    if self.use_keras_api:
        if not glob.glob(os.path.join(model_uri, '*.h5')):
            self.model = tf.keras.models.load_model(model_uri)
        else:
            self.model = tf.keras.models.load_model(glob.glob(os.path.join(model_uri, '*.h5'))[0])
    self.loaded = True
    print(f"Use Keras API: {self.use_keras_api}")
    print(f"Model input layer: {self.model.inputs[0]}")

Predict

def predict(self, X):
    if not self.loaded:
        self.load()
    if self.use_keras_api:
        return self.model.predict(X)
    else:
        output = self.model(tf.convert_to_tensor(X, self.model.inputs[0].dtype))
        return output[next(iter(output))].numpy()

Example

The example uses the Keras MNIST dataset, which is used in tensorflow tutorial.

Property	Description
Model Image	infuseai/tensorflow2-prepackaged:v0.2.0
Model URI	gs://primehub-models/tensorflow2/mnist (SavedModel) or gs://primehub-models/tensorflow2/mnist-h5 (HDF5)

ndarray

Test Request

curl -X POST http://localhost:9000/api/v1.0/predictions \
    -H 'Content-Type: application/json' \
    -d '{ "data": {"ndarray": [[[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.32941176470588235, 0.7254901960784313, 0.6235294117647059, 0.592156862745098, 0.23529411764705882, 0.1411764705882353, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.8705882352941177, 0.996078431372549, 0.996078431372549, 0.996078431372549, 0.996078431372549, 0.9450980392156862, 0.7764705882352941, 0.7764705882352941, 0.7764705882352941, 0.7764705882352941, 0.7764705882352941, 0.7764705882352941, 0.7764705882352941, 0.7764705882352941, 0.6666666666666666, 0.20392156862745098, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.2627450980392157, 0.4470588235294118, 0.2823529411764706, 0.4470588235294118, 0.6392156862745098, 0.8901960784313725, 0.996078431372549, 0.8823529411764706, 0.996078431372549, 0.996078431372549, 0.996078431372549, 0.9803921568627451, 0.8980392156862745, 0.996078431372549, 0.996078431372549, 0.5490196078431373, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.06666666666666667, 0.25882352941176473, 0.054901960784313725, 0.2627450980392157, 0.2627450980392157, 0.2627450980392157, 0.23137254901960785, 0.08235294117647059, 0.9254901960784314, 0.996078431372549, 0.41568627450980394, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.3254901960784314, 0.9921568627450981, 0.8196078431372549, 0.07058823529411765, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.08627450980392157, 0.9137254901960784, 1.0, 0.3254901960784314, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.5058823529411764, 0.996078431372549, 0.9333333333333333, 0.17254901960784313, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.23137254901960785, 0.9764705882352941, 0.996078431372549, 0.24313725490196078, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.5215686274509804, 0.996078431372549, 0.7333333333333333, 0.0196078431372549, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.03529411764705882, 0.803921568627451, 0.9725490196078431, 0.22745098039215686, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.49411764705882355, 0.996078431372549, 0.7137254901960784, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.29411764705882354, 0.984313725490196, 0.9411764705882353, 0.2235294117647059, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.07450980392156863, 0.8666666666666667, 0.996078431372549, 0.6509803921568628, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.011764705882352941, 0.796078431372549, 0.996078431372549, 0.8588235294117647, 0.13725490196078433, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.14901960784313725, 0.996078431372549, 0.996078431372549, 0.30196078431372547, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.12156862745098039, 0.8784313725490196, 0.996078431372549, 0.45098039215686275, 0.00392156862745098, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.5215686274509804, 0.996078431372549, 0.996078431372549, 0.20392156862745098, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.23921568627450981, 0.9490196078431372, 0.996078431372549, 0.996078431372549, 0.20392156862745098, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.4745098039215686, 0.996078431372549, 0.996078431372549, 0.8588235294117647, 0.1568627450980392, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.4745098039215686, 0.996078431372549, 0.8117647058823529, 0.07058823529411765, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0], [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]] } }'

Test Result

{"data":{"names":[],"ndarray":[[2.2179587233495113e-07,1.2331390131237185e-08,2.5685869331937283e-05,0.0001267452462343499,3.6731301333858823e-10,8.802298339105619e-07,1.7313735514723483e-11,0.9998445510864258,5.112421490593988e-07,1.4923105027264683e-06]]},"meta":{"requestPath":{"model":"infuseai/tensorflow2-prepackaged:v0.2.0"}}}

Image

Test Request

curl -F 'binData=@test_image.jpg' http://localhost:9000/api/v1.0/predictions

Test Result

{"data":{"names":[],"tensor":{"shape":[1,10],"values":[2.240761034499883e-07,1.2446706776358951e-08,2.6079718736582436e-05,0.00012795037764590234,3.6888223031716905e-10,8.873528258845909e-07,1.7562255469338872e-11,0.9998427629470825,5.136774916536524e-07,1.4995322317190585e-06]}},"meta":{"requestPath":{"model":"infuseai/tensorflow2-prepackaged:v0.2.0"}}}