mllite / keras2sql Goto Github PK

Sample use case : simple convnet on the MNIST dataset

keras example : https://github.com/keras-team/keras/blob/master/examples/mnist_cnn.py

used layers and activation functions :

model.add(Conv2D(32, kernel_size=(3, 3),
                 activation='relu',
                 input_shape=input_shape))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))

Models with SimpleRNN layer. Use recursive CTE to implement SimpleRNNCell.

Models with Gated Recurrent Unit (GRU) layer. Use recursive CTE to implement GRUCell.

The straightforward use case of keras2sql is to send a pickled model the web service to generate the SQL code.

Something prevented keras models to be picklable. An issue has been created (keras-team/keras#10475). It now fixed. Thanks to the keras team (@farizrahman4u).

The simple demo script should demo some layers (dense + ) and some activation functions (relu + softmax) on the iris dataset.

This is probably a way to avoid some unnecessarily large models and generate an (almost) equivalent SQL code.

Models with Long Short Term Memory layer. Use recursive CTE to implement LSTMCell.

The notebooks included here were created before pickle support was added to Keras

https://github.com/antoinecarme/keras2sql/tree/master/doc

The need to be adapted to send pickled models to the web service (standard use case).