Create a simple MLP for testing Keras model import. Run Keras mnist_mlp_constraints.py example and then save that model and its outputs to disk. ''' from __future__ import print_function import imp import keras.backend as K from util import save_model_details, save_model_output SCRIPT_PATH = '../examples/mnist_mlp_constraints.py' KERAS_VERSION = '_keras_2' PREFIX = 'mnist_mlp_' + K.image_dim_ordering() + KERAS_VERSION OUT_DIR = '.' print('Entering Keras script') example = imp.load_source('example', SCRIPT_PATH) print('Saving model details') save_model_details(example.model, prefix=PREFIX, out_dir=OUT_DIR) print('Saving model outputs') save_model_output(example.model, example.X_test, example.Y_test, nb_examples=100, prefix=PREFIX, out_dir=OUT_DIR) print('DONE!')
activation="tanh", use_bias=True, input_shape=input_shape)) #model.add(Conv1D(name="conv1", filters=3, kernel_size=k, strides=s, padding="causal", data_format=f, dilation_rate=d, activation=None, use_bias=False)) #model.add(GlobalAveragePooling1D(data_format=f)) #model.add(Activation('softmax')) opt = keras.optimizers.RMSprop(learning_rate=0.0001, decay=1e-6) model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy']) print('Saving model details') save_model_details(model, prefix=name, out_dir=OUT_DIR) exp_out = model.predict(features) print("Input = ", str(features.shape), ", Out = ", str(exp_out.shape), " case - ", name) print('Saving model outputs') save_model_output(model, features, exp_out, nb_examples=None, prefix=name, out_dir=OUT_DIR, labels=None) #labels)