def test_serialization():
spark_model = SparkMLlibModel(model,
frequency='epoch',
mode='synchronous',
num_workers=2)
spark_model.save("test.h5")
load_spark_model("test.h5")
def test_serialization(classification_model):
rms = RMSprop()
classification_model.compile(rms, 'categorical_crossentropy', ['acc'])
spark_model = SparkMLlibModel(
classification_model, frequency='epoch', mode='synchronous', num_workers=2)
spark_model.save("test.h5")
loaded_model = load_spark_model("test.h5")
assert loaded_model.master_network.to_yaml()
def test_mllib_model(spark_context):
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(spark_context, x_train, y_train, categorical=True)
# Initialize SparkModel from Keras model and Spark context
spark_model = SparkMLlibModel(model=model, frequency='epoch', mode='synchronous')
# Train Spark model
spark_model.fit(lp_rdd, epochs=5, batch_size=32, verbose=0,
validation_split=0.1, categorical=True, nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
score = spark_model.master_network.evaluate(x_test, y_test, verbose=2)
print('Test accuracy:', score[1])
def test_mllib_model(spark_context, classification_model, mnist_data):
rms = RMSprop()
classification_model.compile(rms, 'categorical_crossentropy', ['acc'])
x_train, y_train, x_test, y_test = mnist_data
x_train = x_train[:1000]
y_train = y_train[:1000]
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(spark_context, x_train,
y_train, categorical=True)
# Initialize SparkModel from tensorflow.keras model and Spark context
spark_model = SparkMLlibModel(
model=classification_model, frequency='epoch', mode='synchronous')
# Train Spark model
spark_model.fit(lp_rdd, epochs=5, batch_size=32, verbose=0,
validation_split=0.1, categorical=True, nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
score = spark_model.master_network.evaluate(x_test, y_test, verbose=2)
assert score
model.add(Activation('softmax'))
# Compile model
rms = RMSprop()
model.compile(rms, "categorical_crossentropy", ['acc'])
# Create Spark context
conf = SparkConf().setAppName('Mnist_Spark_MLP').setMaster('local[8]')
sc = SparkContext(conf=conf)
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(sc, x_train, y_train, categorical=True)
# Initialize SparkModel from tensorflow.keras model and Spark context
spark_model = SparkMLlibModel(model=model,
frequency='epoch',
mode='synchronous')
# Train Spark model
spark_model.fit(lp_rdd,
epochs=5,
batch_size=32,
verbose=0,
validation_split=0.1,
categorical=True,
nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
score = spark_model.master_network.evaluate(x_test, y_test, verbose=2)
print('Test accuracy:', score[1])
model.add(Dropout(0.2))
model.add(Dense(128))
model.add(Activation('relu'))
model.add(Dropout(0.2))
model.add(Dense(10))
model.add(Activation('softmax'))
# Compile model
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=["accuracy"])
# Create Spark context
conf = SparkConf().setAppName('Mnist_Spark_MLP').setMaster('local[8]')
#sc = SparkContext(conf=conf)
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(sc, x_train, y_train, categorical=True)
rdd = lp_to_simple_rdd(lp_rdd, True, nb_classes)
# Initialize SparkModel from Keras model and Spark context
adagrad = elephas_optimizers.Adagrad()
spark_model = SparkMLlibModel(sc, model, optimizer=adagrad, frequency='batch', mode='asynchronous', num_workers=4)
# Train Spark model
spark_model.train(lp_rdd, nb_epoch=nb_epoch, batch_size=batch_size, verbose=0,
validation_split=0.1, categorical=True, nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
loss, acc = spark_model.master_network.evaluate(x_test, y_test, verbose=2)
print('Test accuracy:', acc)
model.add(Activation('relu'))
model.add(Dropout(0.2))
model.add(Dense(10))
model.add(Activation('softmax'))
# Compile model
rms = RMSprop()
model.compile(loss='categorical_crossentropy', optimizer=rms)
# Create Spark context
conf = SparkConf().setAppName('Mnist_Spark_MLP').setMaster('local[8]')
sc = SparkContext(conf=conf)
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(sc, X_train, Y_train, categorical=True)
print(lp_rdd.first())
rdd = lp_to_simple_rdd(lp_rdd, True, nb_classes)
rdd = rdd.repartition(4)
rdd.first()
# Initialize SparkModel from Keras model and Spark context
adadelta = elephas_optimizers.Adadelta()
spark_model = SparkMLlibModel(sc,model, optimizer=adadelta, frequency='batch', mode='asynchronous', num_workers=2)
# Train Spark model
spark_model.train(lp_rdd, nb_epoch=20, batch_size=32, verbose=0, validation_split=0.1, categorical=True, nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
score = spark_model.get_network().evaluate(X_test, Y_test, show_accuracy=True, verbose=2)
print('Test accuracy:', score[1])
rms = RMSprop()
# Create Spark context
conf = SparkConf().setAppName('Mnist_Spark_MLP').setMaster('local[8]')
sc = SparkContext(conf=conf)
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(sc, x_train, y_train, categorical=True)
rdd = lp_to_simple_rdd(lp_rdd, True, nb_classes)
# Initialize SparkModel from Keras model and Spark context
adadelta = elephas_optimizers.Adadelta()
spark_model = SparkMLlibModel(sc,
model,
optimizer=adadelta,
frequency='batch',
mode='asynchronous',
num_workers=2,
master_optimizer=rms)
# Train Spark model
spark_model.train(lp_rdd,
nb_epoch=20,
batch_size=32,
verbose=0,
validation_split=0.1,
categorical=True,
nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
score = spark_model.master_network.evaluate(x_test, y_test, verbose=2)
outputCol="tf_features",
vocabSize=input_dim)
# IDF
idf = sf.IDF(inputCol="tf_features", outputCol="features")
label_string = sf.StringIndexer(inputCol="first_label", outputCol="label")
pipeline_dl = Pipeline(stages=[cv, idf, label_string])
df = pipeline_dl.fit(training_set).transform(training_set)
df = df.rdd.map(lambda x: (LabeledPoint(x[
'label'], MLLibVectors.fromML(x['features']))))
logger.info("Pipeline created ...")
logger.info("Transforms the text into tf idf RDD ...")
model = create_keras_model(input_dim, output_dim)
logger.info("Starts Training ...")
spark_model = SparkMLlibModel(model=model,
frequency='epoch',
mode='asynchronous',
parameter_server_mode='socket')
spark_model.fit(df,
epochs=epochs,
batch_size=132,
verbose=1,
validation_split=0.2,
categorical=True,
nb_classes=output_dim)
logger.info("Training done")
spark_model._master_network.save(save_dir + model_dir + "/" + filename)
logger.info("Program ended succesfully ! Find the model at :" + save_dir +
model_dir + "/" + filename)
# Create Spark context
conf = SparkConf().setAppName('Mnist_Spark_MLP').setMaster('local[8]')
sc = SparkContext(conf=conf)
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(sc, X_train, Y_train, categorical=True)
print(lp_rdd.first())
rdd = lp_to_simple_rdd(lp_rdd, True, nb_classes)
rdd = rdd.repartition(4)
rdd.first()
# Initialize SparkModel from Keras model and Spark context
adadelta = elephas_optimizers.Adadelta()
spark_model = SparkMLlibModel(sc,
model,
optimizer=adadelta,
frequency='batch',
mode='asynchronous',
num_workers=2)
# Train Spark model
spark_model.train(lp_rdd,
nb_epoch=20,
batch_size=32,
verbose=0,
validation_split=0.1,
categorical=True,
nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
score = spark_model.get_network().evaluate(X_test,
Y_test,
metrics=["accuracy"])
# Create Spark context
conf = SparkConf().setAppName('Mnist_Spark_MLP').setMaster('local[8]')
#sc = SparkContext(conf=conf)
# Build RDD from numpy features and labels
lp_rdd = to_labeled_point(sc, x_train, y_train, categorical=True)
rdd = lp_to_simple_rdd(lp_rdd, True, nb_classes)
# Initialize SparkModel from Keras model and Spark context
adagrad = elephas_optimizers.Adagrad()
spark_model = SparkMLlibModel(sc,
model,
optimizer=adagrad,
frequency='batch',
mode='asynchronous',
num_workers=4)
# Train Spark model
spark_model.train(lp_rdd,
nb_epoch=nb_epoch,
batch_size=batch_size,
verbose=0,
validation_split=0.1,
categorical=True,
nb_classes=nb_classes)
# Evaluate Spark model by evaluating the underlying model
loss, acc = spark_model.master_network.evaluate(x_test, y_test, verbose=2)
print('Test accuracy:', acc)
