def test_accuracy_for_single_center(self):
"""Test that parameters obtained are correct for a single center."""
centers, batches = self.streamingKMeansDataGenerator(batches=5,
numPoints=5,
k=1,
d=5,
r=0.1,
seed=0)
stkm = StreamingKMeans(1)
stkm.setInitialCenters([[0., 0., 0., 0., 0.]], [0.])
input_stream = self.ssc.queueStream(
[self.sc.parallelize(batch, 1) for batch in batches])
stkm.trainOn(input_stream)
self.ssc.start()
def condition():
self.assertEqual(stkm.latestModel().clusterWeights, [25.0])
return True
eventually(condition, catch_assertions=True)
realCenters = array_sum(array(centers), axis=0)
for i in range(5):
modelCenters = stkm.latestModel().centers[0][i]
self.assertAlmostEqual(centers[0][i], modelCenters, 1)
self.assertAlmostEqual(realCenters[i], modelCenters, 1)
def test_parameter_convergence(self):
"""Test that the model parameters improve with streaming data."""
slr = StreamingLinearRegressionWithSGD(stepSize=0.2, numIterations=25)
slr.setInitialWeights([0.0])
# Create ten batches with 100 sample points in each.
batches = []
for i in range(10):
batch = LinearDataGenerator.generateLinearInput(
0.0, [10.0], [0.0], [1.0 / 3.0], 100, 42 + i, 0.1)
batches.append(self.sc.parallelize(batch))
model_weights = []
input_stream = self.ssc.queueStream(batches)
input_stream.foreachRDD(
lambda x: model_weights.append(slr.latestModel().weights[0]))
slr.trainOn(input_stream)
self.ssc.start()
def condition():
self.assertEqual(len(model_weights), len(batches))
return True
# We want all batches to finish for this test.
eventually(condition, 90, catch_assertions=True)
w = array(model_weights)
diff = w[1:] - w[:-1]
self.assertTrue(all(diff >= -0.1))
def test_prediction(self):
"""Test prediction on a model with weights already set."""
# Create a model with initial Weights equal to coefs
slr = StreamingLinearRegressionWithSGD(stepSize=0.2, numIterations=25)
slr.setInitialWeights([10.0, 10.0])
# Create ten batches with 100 sample points in each.
batches = []
for i in range(10):
batch = LinearDataGenerator.generateLinearInput(
0.0, [10.0, 10.0], [0.0, 0.0], [1.0 / 3.0, 1.0 / 3.0], 100,
42 + i, 0.1)
batches.append(
self.sc.parallelize(batch).map(lambda lp:
(lp.label, lp.features)))
input_stream = self.ssc.queueStream(batches)
output_stream = slr.predictOnValues(input_stream)
samples = []
output_stream.foreachRDD(lambda x: samples.append(x.collect()))
self.ssc.start()
def condition():
self.assertEqual(len(samples), len(batches))
return True
# We want all batches to finish for this test.
eventually(condition, catch_assertions=True)
# Test that mean absolute error on each batch is less than 0.1
for batch in samples:
true, predicted = zip(*batch)
self.assertTrue(mean(abs(array(true) - array(predicted))) < 0.1)
def test_predictions(self):
"""Test predicted values on a toy model."""
input_batches = []
for i in range(20):
batch = self.sc.parallelize(
self.generateLogisticInput(0, 1.5, 100, 42 + i))
input_batches.append(batch.map(lambda x: (x.label, x.features)))
input_stream = self.ssc.queueStream(input_batches)
slr = StreamingLogisticRegressionWithSGD(stepSize=0.2,
numIterations=25)
slr.setInitialWeights([1.5])
predict_stream = slr.predictOnValues(input_stream)
true_predicted = []
predict_stream.foreachRDD(lambda x: true_predicted.append(x.collect()))
self.ssc.start()
def condition():
self.assertEqual(len(true_predicted), len(input_batches))
return True
eventually(condition, catch_assertions=True)
# Test that the accuracy error is no more than 0.4 on each batch.
for batch in true_predicted:
true, predicted = zip(*batch)
self.assertTrue(
self.calculate_accuracy_error(true, predicted) < 0.4)
def test_parameter_accuracy(self):
"""Test that coefs are predicted accurately by fitting on toy data."""
# Test that fitting (10*X1 + 10*X2), (X1, X2) gives coefficients
# (10, 10)
slr = StreamingLinearRegressionWithSGD(stepSize=0.2, numIterations=25)
slr.setInitialWeights([0.0, 0.0])
xMean = [0.0, 0.0]
xVariance = [1.0 / 3.0, 1.0 / 3.0]
# Create ten batches with 100 sample points in each.
batches = []
for i in range(10):
batch = LinearDataGenerator.generateLinearInput(
0.0, [10.0, 10.0], xMean, xVariance, 100, 42 + i, 0.1)
batches.append(self.sc.parallelize(batch))
input_stream = self.ssc.queueStream(batches)
slr.trainOn(input_stream)
self.ssc.start()
def condition():
self.assertArrayAlmostEqual(slr.latestModel().weights.array,
[10., 10.], 1)
self.assertAlmostEqual(slr.latestModel().intercept, 0.0, 1)
return True
eventually(condition, catch_assertions=True)
def test_trainOn_predictOn(self):
"""Test that prediction happens on the updated model."""
stkm = StreamingKMeans(decayFactor=0.0, k=2)
stkm.setInitialCenters([[0.0], [1.0]], [1.0, 1.0])
# Since decay factor is set to zero, once the first batch
# is passed the clusterCenters are updated to [-0.5, 0.7]
# which causes 0.2 & 0.3 to be classified as 1, even though the
# classification based in the initial model would have been 0
# proving that the model is updated.
batches = [[[-0.5], [0.6], [0.8]], [[0.2], [-0.1], [0.3]]]
batches = [self.sc.parallelize(batch) for batch in batches]
input_stream = self.ssc.queueStream(batches)
predict_results = []
def collect(rdd):
rdd_collect = rdd.collect()
if rdd_collect:
predict_results.append(rdd_collect)
stkm.trainOn(input_stream)
predict_stream = stkm.predictOn(input_stream)
predict_stream.foreachRDD(collect)
self.ssc.start()
def condition():
self.assertEqual(predict_results, [[0, 1, 1], [1, 0, 1]])
return True
eventually(condition, catch_assertions=True)
def test_convergence(self):
"""
Test that weights converge to the required value on toy data.
"""
input_batches = [
self.sc.parallelize(self.generateLogisticInput(
0, 1.5, 100, 42 + i)) for i in range(20)
]
input_stream = self.ssc.queueStream(input_batches)
models = []
slr = StreamingLogisticRegressionWithSGD(stepSize=0.2,
numIterations=25)
slr.setInitialWeights([0.0])
slr.trainOn(input_stream)
input_stream.foreachRDD(
lambda x: models.append(slr.latestModel().weights[0]))
self.ssc.start()
def condition():
self.assertEqual(len(models), len(input_batches))
return True
# We want all batches to finish for this test.
eventually(condition, 60.0, catch_assertions=True)
t_models = array(models)
diff = t_models[1:] - t_models[:-1]
# Test that weights improve with a small tolerance
self.assertTrue(all(diff >= -0.1))
self.assertTrue(array_sum(diff > 0) > 1)
def test_trainOn_model(self):
"""Test the model on toy data with four clusters."""
stkm = StreamingKMeans()
initCenters = [[1.0, 1.0], [-1.0, 1.0], [-1.0, -1.0], [1.0, -1.0]]
stkm.setInitialCenters(centers=initCenters,
weights=[1.0, 1.0, 1.0, 1.0])
# Create a toy dataset by setting a tiny offset for each point.
offsets = [[0, 0.1], [0, -0.1], [0.1, 0], [-0.1, 0]]
batches = []
for offset in offsets:
batches.append([[offset[0] + center[0], offset[1] + center[1]]
for center in initCenters])
batches = [self.sc.parallelize(batch, 1) for batch in batches]
input_stream = self.ssc.queueStream(batches)
stkm.trainOn(input_stream)
self.ssc.start()
# Give enough time to train the model.
def condition():
finalModel = stkm.latestModel()
self.assertTrue(all(finalModel.centers == array(initCenters)))
self.assertEqual(finalModel.clusterWeights, [5.0, 5.0, 5.0, 5.0])
return True
eventually(condition, catch_assertions=True)
def test_predictOn_model(self):
"""Test that the model predicts correctly on toy data."""
stkm = StreamingKMeans()
stkm._model = StreamingKMeansModel(clusterCenters=[[1.0, 1.0],
[-1.0, 1.0],
[-1.0, -1.0],
[1.0, -1.0]],
clusterWeights=[1.0, 1.0, 1.0, 1.0])
predict_data = [[[1.5, 1.5]], [[-1.5, 1.5]], [[-1.5, -1.5]],
[[1.5, -1.5]]]
predict_data = [
self.sc.parallelize(batch, 1) for batch in predict_data
]
predict_stream = self.ssc.queueStream(predict_data)
predict_val = stkm.predictOn(predict_stream)
result = []
def update(rdd):
rdd_collect = rdd.collect()
if rdd_collect:
result.append(rdd_collect)
predict_val.foreachRDD(update)
self.ssc.start()
def condition():
self.assertEqual(result, [[0], [1], [2], [3]])
return True
eventually(condition, catch_assertions=True)
def test_train_prediction(self):
"""Test that error on test data improves as model is trained."""
slr = StreamingLinearRegressionWithSGD(stepSize=0.2, numIterations=25)
slr.setInitialWeights([0.0])
# Create fifteen batches with 100 sample points in each.
batches = []
for i in range(15):
batch = LinearDataGenerator.generateLinearInput(
0.0, [10.0], [0.0], [1.0 / 3.0], 100, 42 + i, 0.1)
batches.append(self.sc.parallelize(batch))
predict_batches = [
b.map(lambda lp: (lp.label, lp.features)) for b in batches]
errors = []
def func(rdd):
true, predicted = zip(*rdd.collect())
errors.append(mean(abs(true) - abs(predicted)))
input_stream = self.ssc.queueStream(batches)
output_stream = self.ssc.queueStream(predict_batches)
slr.trainOn(input_stream)
output_stream = slr.predictOnValues(output_stream)
output_stream.foreachRDD(func)
self.ssc.start()
def condition():
if len(errors) == len(predict_batches):
self.assertGreater(errors[1] - errors[-1], 2)
if len(errors) >= 3 and errors[1] - errors[-1] > 2:
return True
return "Latest errors: " + ", ".join(map(lambda x: str(x), errors))
eventually(condition, timeout=180.0)
def test_training_and_prediction(self):
"""Test that the model improves on toy data with no. of batches"""
input_batches = [
self.sc.parallelize(self.generateLogisticInput(0, 1.5, 100, 42 + i))
for i in range(40)]
predict_batches = [
b.map(lambda lp: (lp.label, lp.features)) for b in input_batches]
slr = StreamingLogisticRegressionWithSGD(
stepSize=0.01, numIterations=25)
slr.setInitialWeights([-0.1])
errors = []
def collect_errors(rdd):
true, predicted = zip(*rdd.collect())
errors.append(self.calculate_accuracy_error(true, predicted))
true_predicted = []
input_stream = self.ssc.queueStream(input_batches)
predict_stream = self.ssc.queueStream(predict_batches)
slr.trainOn(input_stream)
ps = slr.predictOnValues(predict_stream)
ps.foreachRDD(lambda x: collect_errors(x))
self.ssc.start()
def condition():
# Test that the improvement in error is > 0.3
if len(errors) == len(predict_batches):
self.assertGreater(errors[1] - errors[-1], 0.3)
if len(errors) >= 3 and errors[1] - errors[-1] > 0.3:
return True
return "Latest errors: " + ", ".join(map(lambda x: str(x), errors))
eventually(condition, timeout=180.0)
def test_reuse_worker_of_parallelize_range(self):
def check_reuse_worker_of_parallelize_range():
rdd = self.sc.parallelize(range(20), 8)
previous_pids = rdd.map(lambda x: os.getpid()).collect()
current_pids = rdd.map(lambda x: os.getpid()).collect()
for pid in current_pids:
self.assertTrue(pid in previous_pids)
return True
eventually(check_reuse_worker_of_parallelize_range, catch_assertions=True)
def test_java_object_gets_detached(self):
df = self.spark.createDataFrame(
[(1.0, 2.0, Vectors.dense(1.0)),
(0.0, 2.0, Vectors.sparse(1, [], []))],
["label", "weight", "features"],
)
lr = LinearRegression(maxIter=1,
regParam=0.0,
solver="normal",
weightCol="weight",
fitIntercept=False)
model = lr.fit(df)
summary = model.summary
self.assertIsInstance(model, JavaWrapper)
self.assertIsInstance(summary, JavaWrapper)
self.assertIsInstance(model, JavaParams)
self.assertNotIsInstance(summary, JavaParams)
error_no_object = "Target Object ID does not exist for this gateway"
self.assertIn("LinearRegression_", model._java_obj.toString())
self.assertIn("LinearRegressionTrainingSummary",
summary._java_obj.toString())
model.__del__()
def condition():
with self.assertRaisesRegex(py4j.protocol.Py4JError,
error_no_object):
model._java_obj.toString()
self.assertIn("LinearRegressionTrainingSummary",
summary._java_obj.toString())
return True
eventually(condition, timeout=10, catch_assertions=True)
try:
summary.__del__()
except:
pass
def condition():
with self.assertRaisesRegex(py4j.protocol.Py4JError,
error_no_object):
model._java_obj.toString()
with self.assertRaisesRegex(py4j.protocol.Py4JError,
error_no_object):
summary._java_obj.toString()
return True
eventually(condition, timeout=10, catch_assertions=True)
def test_parameter_accuracy(self):
"""
Test that the final value of weights is close to the desired value.
"""
input_batches = [
self.sc.parallelize(self.generateLogisticInput(0, 1.5, 100, 42 + i)) for i in range(20)
]
input_stream = self.ssc.queueStream(input_batches)
slr = StreamingLogisticRegressionWithSGD(stepSize=0.2, numIterations=25)
slr.setInitialWeights([0.0])
slr.trainOn(input_stream)
self.ssc.start()
def condition():
rel = (1.5 - slr.latestModel().weights.array[0]) / 1.5
self.assertAlmostEqual(rel, 0.1, 1)
return True
eventually(condition, timeout=120.0, catch_assertions=True)
def test_task_context_correct_with_python_worker_reuse(self):
# Retrying the check as the PIDs from Python workers might be different even
# when reusing Python workers is enabled if a Python worker is dead for some reasons
# (e.g., socket connection failure) and new Python worker is created.
eventually(
self.check_task_context_correct_with_python_worker_reuse, catch_assertions=True)