def __init__(self, model_name, model_base_path, verbose=False):
"""
Initialize the service.
Args:
model_name: The name of the model.
model_base_path: The file path of the model.
Return:
None
"""
super(PmmlInferenceService, self).__init__()
self.model_name = model_name
self.model_base_path = model_base_path
self.model_version_list = [1]
self.model_graph_signature = ""
self.platform = "PMML"
# Load model
from openscoring import Openscoring
openscoring_server_endpoint = "localhost:8080"
kwargs = {"auth" : ("admin", "adminadmin")}
self.openscoring = Openscoring("http://{}/openscoring".format(openscoring_server_endpoint))
self.openscoring.deployFile(self.model_name, self.model_base_path, **kwargs)
self.model_graph_signature = "No signature for PMML models"
self.verbose = verbose
def testMergeDictOverride(self):
self.assertEqual({"A": {
"one": 1
}}, Openscoring._merge({"A": {
"one": 1
}}))
self.assertEqual({"A": {
"one": "1"
}}, Openscoring._merge({"A": {
"one": 1
}}, A={"one": "1"}))
class FNDOpenScoring:
def __init__(self):
self.scoring = Openscoring("http://localhost:8080/openscoring")
self.arguments = {"sourceTaggedAsFakeCount": 2.0, "reporterScore": 3.0}
def _deploy(self):
self.scoring.deployFile("FakeNewsDetector", "FakeNewsAIModel.pmml",
**{"auth": ("admin", "adminadmin")})
def _setArgs(self, jsonString):
jsonObj = json.loads(jsonString)
self.arguments["similarWebAvgScore"] = jsonObj['average']
self.arguments["similarWebStdScore"] = jsonObj['variance']
self.arguments["fakeFactSitesCount"] = jsonObj['blacklisted']
def __call__(self, jsonString):
self._deploy()
self._setArgs(jsonString)
result = self.scoring.evaluate("FakeNewsDetector", self.arguments)
return result
def testMergeDict(self):
self.assertEqual({"A": {
"one": 1,
"two": 2,
"three": 3
}}, Openscoring._merge({"A": {
"one": 1
}}, A={
"two": 2,
"three": 3
}))
def get_predictions(pw, pl, sw, sl):
"""
Get predictions by source
:param pw:
:param pl:
:param sw:
:param sl:
:return:
"""
arguments = {
"sepal_length": float(sl),
"sepal_width": float(sw),
"petal_length": float(pl),
"petal_width": float(pw)
}
print(arguments)
from openscoring import Openscoring
os = Openscoring("http://localhost:8080/openscoring")
result = os.evaluate("Iris", arguments)
print(result)
return html.Div([
html.H3("The class detected was : {}".format(result['species'])),
])
def testMergeDict(self):
self.assertEqual({"A" : {"one" : 1, "two" : 2, "three" : 3}}, Openscoring._merge({"A" : {"one" : 1}}, A = {"two" : 2, "three" : 3}))
def testMergeValueConflict(self):
with self.assertRaises(Exception):
Openscoring._merge({"A" : 1}, A = "1")
def testMissingUserDict(self):
self.assertEqual({}, Openscoring._merge(None))
self.assertEqual({"A": 1}, Openscoring._merge(None, A=1))
self.assertEqual({"A": {
"one": 1
}}, Openscoring._merge(None, A={"one": 1}))
#!/usr/bin/env python
from openscoring import Openscoring
os = Openscoring("http://localhost:8080/openscoring")
kwargs = {"auth": ("admin", "adminadmin")}
model_file_path = "../../models/pmml_iris/DecisionTreeIris.pmml"
model_name = "PmmlModel"
os.deployFile(model_name, model_file_path, **kwargs)
arguments = {
"Sepal_Length": 5.1,
"Sepal_Width": 3.5,
"Petal_Length": 1.4,
"Petal_Width": 0.2
}
result = os.evaluate(model_name, arguments)
print(result)
app.config['suppress_callback_exceptions'] = True
if __name__ == '__main__':
import pandas
from sklearn.tree import DecisionTreeClassifier
from sklearn2pmml.pipeline import PMMLPipeline
from sklearn2pmml import sklearn2pmml
iris_df = pandas.read_csv("iris.csv")
#
pipeline = PMMLPipeline([("classifier", DecisionTreeClassifier())])
pipeline.fit(iris_df[iris_df.columns.difference(["species"])],
iris_df["species"])
#
sklearn2pmml(pipeline, "DecisionTreeIris.pmml", with_repr=True)
## Actual openscoring code
from openscoring import Openscoring
import subprocess
p = subprocess.Popen('java -jar openscoring-server-executable-1.4.3.jar',
shell=True)
os = Openscoring("http://localhost:8080/openscoring")
#
kwargs = {"auth": ("admin", "adminadmin")}
os.deployFile("Iris", "DecisionTreeIris.pmml", **kwargs)
app.run_server(debug=True, port=9000)
("Income", ContinuousDomain()),
(["Hours", "Income"],
Alias(ExpressionTransformer("X[1] / (X[0] * 52)"),
"Hourly_Income"))])
classifier = H2ORandomForestEstimator(ntrees=17)
predict_proba_transformer = Pipeline([
("expression", ExpressionTransformer("X[1]")),
("cut",
Alias(CutTransformer(bins=[0.0, 0.75, 0.90, 1.0],
labels=["no", "maybe", "yes"]),
"Decision",
prefit=True))
])
pipeline = PMMLPipeline([("local_mapper", mapper),
("uploader", H2OFrameCreator()),
("remote_classifier", classifier)],
predict_proba_transformer=predict_proba_transformer)
pipeline.fit(audit_X, H2OFrame(audit_y.to_frame(),
column_types=["categorical"]))
pipeline.verify(audit_X.sample(100))
sklearn2pmml(pipeline, "pmml/RandomForestAudit.pmml")
if "--deploy" in sys.argv:
from openscoring import Openscoring
os = Openscoring("http://localhost:8080/openscoring")
os.deployFile("RandomForestAudit", "pmml/RandomForestAudit.pmml")
def testMergeValueConflict(self):
with self.assertRaises(Exception):
Openscoring._merge({"A": 1}, A="1")
def testMergeValueEqual(self):
self.assertEqual({"A": 1}, Openscoring._merge({"A": 1}, A=1))
def testMergeValue(self):
self.assertEqual({
"A": 1,
"B": 2,
"C": 3
}, Openscoring._merge({"A": 1}, B=2, C=3))
def testMergeDictOverride(self):
self.assertEqual({"A" : {"one" : 1}}, Openscoring._merge({"A" : {"one" : 1}}))
self.assertEqual({"A" : {"one" : "1"}}, Openscoring._merge({"A" : {"one" : 1}}, A = {"one" : "1"}))
def testMissingUserDict(self):
self.assertEqual({}, Openscoring._merge(None))
self.assertEqual({"A" : 1}, Openscoring._merge(None, A = 1))
self.assertEqual({"A" : {"one" : 1}}, Openscoring._merge(None, A = {"one" : 1}))
LabelBinarizer()
]), ("Hours", ContinuousDomain()), ("Income", ContinuousDomain()),
(["Hours", "Income"],
Alias(ExpressionTransformer("X[1] / (X[0] * 52)"), "Hourly_Income"))
])
interaction_mapper = DataFrameMapper([
("Gender", [CategoricalDomain(), LabelBinarizer()]),
("Marital", [CategoricalDomain(), LabelBinarizer()])
])
classifier = XGBClassifier()
pipeline = PMMLPipeline([
("mapper",
FeatureUnion([("scalar_mapper", scalar_mapper),
("interaction",
Pipeline([("interaction_mapper", interaction_mapper),
("polynomial", PolynomialFeatures())]))])),
("classifier", classifier)
])
pipeline.fit(audit_X, audit_y)
pipeline.configure(compact=True)
pipeline.verify(audit_X.sample(100), zeroThreshold=1e-6, precision=1e-6)
sklearn2pmml(pipeline, "pmml/XGBoostAudit.pmml")
if "--deploy" in sys.argv:
from openscoring import Openscoring
os = Openscoring("http://localhost:8080/openscoring")
os.deployFile("XGBoostAudit", "pmml/XGBoostAudit.pmml")
from sklearn2pmml import sklearn2pmml
from sklearn2pmml.pipeline import PMMLPipeline
from sklearn2pmml.ruleset import RuleSetClassifier
import pandas
import sys
iris_df = pandas.read_csv("csv/Iris.csv")
#print(iris_df.head(5))
iris_X = iris_df[iris_df.columns.difference(["Species"])]
iris_y = iris_df["Species"]
classifier = RuleSetClassifier([
("X['Petal_Length'] < 2.45", "setosa"),
("X['Petal_Width'] < 1.75", "versicolor"),
], default_score = "virginica")
pipeline = PMMLPipeline([
("classifier", classifier)
])
pipeline.fit(iris_X, iris_y)
sklearn2pmml(pipeline, "pmml/RuleSetIris.pmml")
if "--deploy" in sys.argv:
from openscoring import Openscoring
os = Openscoring("http://localhost:8080/openscoring")
os.deployFile("RuleSetIris", "pmml/RuleSetIris.pmml")
def testReadme(self):
openscoring = Openscoring()
pmml = os.path.join(os.path.dirname(__file__), "resources",
"DecisionTreeIris.pmml")
with open(pmml, "rb") as instream:
pmmlBytes = instream.read()
self.assertTrue(isinstance(pmmlBytes, bytes))
self.assertEqual(4306, len(pmmlBytes))
modelResponse = openscoring.deploy("Iris", pmmlBytes)
self.assertEqual("Iris", modelResponse.id)
modelResponse = openscoring.deployFile("Iris", pmml)
self.assertEqual("Iris", modelResponse.id)
arguments = {
"Sepal_Length": 5.1,
"Sepal_Width": 3.5,
"Petal_Length": 1.4,
"Petal_Width": 0.2
}
result = openscoring.evaluate("Iris", arguments)
self.assertEqual(
{
"Species": "setosa",
"Probability_setosa": 1.0,
"Probability_versicolor": 0.0,
"Probability_virginica": 0.0,
"Node_Id": "2"
}, result)
evaluationRequest = EvaluationRequest("record-001", arguments)
evaluationResponse = openscoring.evaluate("Iris", evaluationRequest)
self.assertEqual(evaluationRequest.id, evaluationResponse.id)
self.assertEqual("setosa", evaluationResponse.result["Species"])
inCsv = os.path.join(os.path.dirname(__file__), "resources",
"input.csv")
outCsv = os.path.join(tempfile.gettempdir(), "output.csv")
df = pandas.read_csv(inCsv, sep=",")
result = openscoring.evaluateCsv("Iris", df)
self.assertEqual(df["Id"].tolist(), result["Id"].tolist())
self.assertEqual(["setosa", "versicolor", "virginica"],
result["Species"].tolist())
self.assertFalse(os.path.isfile(outCsv))
openscoring.evaluateCsvFile("Iris", inCsv, outCsv)
self.assertTrue(
os.path.isfile(outCsv) and os.path.getsize(outCsv) > 10)
os.remove(outCsv)
openscoring.undeploy("Iris")
with self.assertRaises(Exception) as context:
openscoring.evaluate("Iris", arguments)
self.assertEqual("Model \"Iris\" not found", str(context.exception))
class PmmlInferenceService(AbstractInferenceService):
"""
The service to load PMML model and make inference.
"""
def __init__(self, model_name, model_base_path, verbose=False):
"""
Initialize the service.
Args:
model_name: The name of the model.
model_base_path: The file path of the model.
Return:
None
"""
super(PmmlInferenceService, self).__init__()
self.model_name = model_name
self.model_base_path = model_base_path
self.model_version_list = [1]
self.model_graph_signature = ""
self.platform = "PMML"
# Load model
from openscoring import Openscoring
openscoring_server_endpoint = "localhost:8080"
kwargs = {"auth" : ("admin", "adminadmin")}
self.openscoring = Openscoring("http://{}/openscoring".format(openscoring_server_endpoint))
self.openscoring.deployFile(self.model_name, self.model_base_path, **kwargs)
self.model_graph_signature = "No signature for PMML models"
self.verbose = verbose
def inference(self, json_data):
"""
Make inference with the current Session object and JSON request data.
Args:
json_data: The JSON serialized object with key and array data.
Example is {"model_version": 1, "data": {"keys": [[1.0], [2.0]], "features": [[10, 10, 10, 8, 6, 1, 8, 9, 1], [6, 2, 1, 1, 1, 1, 7, 1, 1]]}}.
Return:
The dictionary with key and array data.
Example is {"keys": [[11], [2]], "softmax": [[0.61554497, 0.38445505], [0.61554497, 0.38445505]], "prediction": [0, 0]}.
"""
# 1. Build inference data
# Example: arguments = {"Sepal_Length" : 5.1, "Sepal_Width" : 3.5, "Petal_Length" : 1.4, "Petal_Width" : 0.2}
request_json_data = json_data["data"]
# 2. Do inference
if self.verbose:
start_time = time.time()
# Example: {'Probability_setosa': 1.0, 'Probability_versicolor': 0.0, 'Node_Id': '2', 'Species': 'setosa', 'Probability_virginica': 0.0}
predict_result = self.openscoring.evaluate(self.model_name, request_json_data)
if self.verbose:
logging.debug("Inference time: {} s".format(time.time() - start_time))
# 3. Build return data
result = {
"result": predict_result,
}
if self.verbose:
logging.debug("Inference result: {}".format(result))
return result
def testMergeValue(self):
self.assertEqual({"A" : 1, "B" : 2, "C" : 3}, Openscoring._merge({"A" : 1}, B = 2, C = 3))
def testReadme(self):
openscoring = Openscoring(base_url = "http://localhost:8080/openscoring", token = os.getenv("OPENSCORING_TOKEN", None))
pmml = os.path.join(os.path.dirname(__file__), "resources", "DecisionTreeIris.pmml")
with open(pmml, "rb") as instream:
pmmlBytes = instream.read()
self.assertTrue(isinstance(pmmlBytes, bytes))
self.assertEqual(2919, len(pmmlBytes))
modelResponse = openscoring.deploy("Iris", pmmlBytes)
self.assertEqual("Iris", modelResponse.id)
modelResponse = openscoring.deployFile("Iris", pmml)
self.assertEqual("Iris", modelResponse.id)
arguments = {
"Sepal.Length" : 5.1,
"Sepal.Width" : 3.5,
"Petal.Length" : 1.4,
"Petal.Width" : 0.2
}
results = openscoring.evaluate("Iris", arguments)
self.assertEqual({"Species" : "setosa", "probability(setosa)" : 1.0, "probability(versicolor)" : 0.0, "probability(virginica)" : 0.0}, results)
evaluationRequest = EvaluationRequest("record-001", arguments)
evaluationResponse = openscoring.evaluate("Iris", evaluationRequest)
self.assertEqual(evaluationRequest.id, evaluationResponse.id)
self.assertEqual("setosa", evaluationResponse.results["Species"])
batchArguments = [
{
"Petal.Length" : 1.4,
"Petal.Width" : 0.2
},
{
"Petal.Length" : 4.7,
"Petal.Width" : 1.4
},
{
"Petal.Length" : 3.6,
"Petal.Width" : 2.5
}
]
batchResults = openscoring.evaluateBatch("Iris", batchArguments)
self.assertEquals(3, len(batchResults))
self.assertEquals({"Species" : "setosa", "probability(setosa)" : 1.0, "probability(versicolor)" : 0.0, "probability(virginica)" : 0.0}, batchResults[0])
self.assertEquals({"Species" : "versicolor", "probability(setosa)" : 0.0, "probability(versicolor)" : (49.0 / 54.0), "probability(virginica)" : (5.0 / 54.0)}, batchResults[1])
self.assertEquals({"Species" : "virginica", "probability(setosa)" : 0.0, "probability(versicolor)" : (1.0 / 46.0), "probability(virginica)" : (45.0 / 46.0)}, batchResults[2])
evaluationRequests = [EvaluationRequest(None, arguments) for arguments in batchArguments]
batchEvaluationRequest = BatchEvaluationRequest("batch-A", evaluationRequests)
batchEvaluationResponse = openscoring.evaluateBatch("Iris", batchEvaluationRequest)
self.assertEqual(batchEvaluationRequest.id, batchEvaluationResponse.id)
evaluationResponses = batchEvaluationResponse.responses
self.assertEqual(3, len(evaluationResponses))
self.assertEqual("setosa", evaluationResponses[0].results["Species"])
self.assertEqual("versicolor", evaluationResponses[1].results["Species"])
self.assertEqual("virginica", evaluationResponses[2].results["Species"])
inCsv = os.path.join(os.path.dirname(__file__), "resources", "input.csv")
outCsv = os.path.join(tempfile.gettempdir(), "output.csv")
dfArguments = pandas.read_csv(inCsv, sep = ",")
dfResults = openscoring.evaluateCsv("Iris", dfArguments)
self.assertEqual((3, 1 + 4), dfResults.shape)
self.assertEqual(dfArguments["Id"].tolist(), dfResults["Id"].tolist())
self.assertEqual(["setosa", "versicolor", "virginica"], dfResults["Species"].tolist())
self.assertFalse(os.path.isfile(outCsv))
openscoring.evaluateCsvFile("Iris", inCsv, outCsv)
self.assertTrue(os.path.isfile(outCsv) and os.path.getsize(outCsv) > 10)
os.remove(outCsv)
openscoring.undeploy("Iris")
with self.assertRaises(Exception) as context:
openscoring.evaluate("Iris", arguments)
self.assertEqual("Not Found", str(context.exception))
def testMergeValueEqual(self):
self.assertEqual({"A" : 1}, Openscoring._merge({"A" : 1}, A = 1))
def __init__(self):
self.scoring = Openscoring("http://localhost:8080/openscoring")
self.arguments = {"sourceTaggedAsFakeCount": 2.0, "reporterScore": 3.0}
level=logging.DEBUG,
format='%(asctime)s %(message)s')
logging.getLogger('requests').setLevel(logging.NOTSET)
client = mqtt.Client()
client.on_connect = on_connect
client.on_message = on_message
## Callback specification
client.message_callback_add("pmml/allongement", on_message_allongement)
client.message_callback_add("pmml/durete", on_message_durete)
client.message_callback_add("pmml/rupture", on_message_rupture)
client.message_callback_add("pmml/100", on_message_100)
## Openscoring definition and model deployment
os = Openscoring("http://localhost:8080/openscoring")
os.deployFile("rupture", "model/rank_contrainterupture.pmml")
os.deployFile("100", "model/rank_contrainte100.pmml")
os.deployFile("allongement", "model/rank_allongement.pmml")
os.deployFile("durete", "model/rank_durete.pmml")
## Global variable definition
urllib3.contrib.pyopenssl.inject_into_urllib3()
http = urllib3.PoolManager(cert_reqs='CERT_REQUIRED', ca_certs=certifi.where())
reset_data()
batch_consumed = ""
server_rheo = 'tcp:SLVDSQL02'
