PyPMML

PyPMML is a Python PMML scoring library, it really is the Python API for PMML4S.

Prerequisites

• Java >= 8 and < 16
• Python 2.7 or >= 3.5

Dependencies

• Py4J

Installation

pip install pypmml

Or install the latest version from github:

pip install --upgrade git+https://github.com/autodeployai/pypmml.git

Usage

1. Load model from various sources, e.g. readable, file path, string, or an array of bytes.

   from pypmml import Model
   
   # The model is from http://dmg.org/pmml/pmml_examples/KNIME_PMML_4.1_Examples/single_iris_dectree.xml
   model = Model.load('single_iris_dectree.xml')

2. Call predict(data) to predict new values that can be in different types, e.g. dict, list, json, ndarray of NumPy, Series or DataFrame of Pandas.

   • data in dict:

   >>> model.predict({'sepal_length': 5.1, 'sepal_width': 3.5, 'petal_length': 1.4, 'petal_width': 0.2})
   {'probability_Iris-setosa': 1.0, 'probability_Iris-versicolor': 0.0, 'probability': 1.0, 'predicted_class': 'Iris-setosa', 'probability_Iris-virginica': 0.0, 'node_id': '1'}

   • data in list:

   NOTE: the order of values must match the input names, and the order of results always matches the output names.

   >>> model.inputNames
   ['sepal_length', 'sepal_width', 'petal_length', 'petal_width']
   >>> model.predict([5.1, 3.5, 1.4, 0.2])
   ['Iris-setosa', 1.0, 1.0, 0.0, 0.0, '1']
   >>> model.outputNames
   ['predicted_class', 'probability', 'probability_Iris-setosa', 'probability_Iris-versicolor', 'probability_Iris-virginica', 'node_id']

   • data in records json:

   >>> model.predict('[{"sepal_length": 5.1, "sepal_width": 3.5, "petal_length": 1.4, "petal_width": 0.2}]')
   [{"probability":1.0,"probability_Iris-versicolor":0.0,"probability_Iris-setosa":1.0,"probability_Iris-virginica":0.0,"predicted_class":"Iris-setosa","node_id":"1"}]

   • data in split json:

   >>> model.predict('{"columns": ["sepal_length", "sepal_width", "petal_length", "petal_width"], "data": [[5.1, 3.5, 1.4, 0.2]]}')
   {"columns":["predicted_class","probability","probability_Iris-setosa","probability_Iris-versicolor","probability_Iris-virginica","node_id"],"data":[["Iris-setosa",1.0,1.0,0.0,0.0,"1"]]}

   • data in ndarray of NumPy:

   NOTE: as the list above, the order of ndarray values must match the input names, and the order of results always matches the output names.

   >>> import numpy as np
   >>> model.predict(np.array([5.1, 3.5, 1.4, 0.2]))
   ['Iris-setosa', 1.0, 1.0, 0.0, 0.0, '1']
   >>> 
   >>> model.predict(np.array([[5.1, 3.5, 1.4, 0.2], [7, 3.2, 4.7, 1.4]]))
   [['Iris-setosa', 1.0, 1.0, 0.0, 0.0, '1'], ['Iris-versicolor', 0.9074074074074074, 0.0, 0.9074074074074074, 0.09259259259259259, '3']]

   • data in Series of Pandas:

   >>> import pandas as pd
   >>> model.predict(pd.Series({'sepal_length': 5.1, 'sepal_width': 3.5, 'petal_length': 1.4, 'petal_width': 0.2}))
   node_id                                  1
   predicted_class                Iris-setosa
   probability                              1
   probability_Iris-setosa                  1
   probability_Iris-versicolor              0
   probability_Iris-virginica               0
   Name: 0, dtype: object

   • data in DataFrame of Pandas:

   >>> import pandas as pd
   >>> data = pd.read_csv('Iris.csv') # The data is from here: http://dmg.org/pmml/pmml_examples/Iris.csv
   >>> model.predict(data)
   node_id predicted_class  probability  probability_Iris-setosa  probability_Iris-versicolor  probability_Iris-virginica
   0         1     Iris-setosa     1.000000                      1.0                     0.000000                    0.000000
   1         1     Iris-setosa     1.000000                      1.0                     0.000000                    0.000000
   2         1     Iris-setosa     1.000000                      1.0                     0.000000                    0.000000
   3         1     Iris-setosa     1.000000                      1.0                     0.000000                    0.000000
   4         1     Iris-setosa     1.000000                      1.0                     0.000000                    0.000000
   ..      ...             ...          ...                      ...                          ...                         ...
   145      10  Iris-virginica     0.978261                      0.0                     0.021739                    0.978261
   146      10  Iris-virginica     0.978261                      0.0                     0.021739                    0.978261
   147      10  Iris-virginica     0.978261                      0.0                     0.021739                    0.978261
   148      10  Iris-virginica     0.978261                      0.0                     0.021739                    0.978261
   149      10  Iris-virginica     0.978261                      0.0                     0.021739                    0.978261

Use PMML in Scala or Java

See the PMML4S project. PMML4S is a PMML scoring library for Scala. It provides both Scala and Java Evaluator API for PMML.

Use PMML in Spark

See the PMML4S-Spark project. PMML4S-Spark is a PMML scoring library for Spark as SparkML Transformer.

Use PMML in PySpark

See the PyPMML-Spark project. PyPMML-Spark is a Python PMML scoring library for PySpark as SparkML Transformer, it really is the Python API for PMML4s-Spark.

Deploy PMML as REST API

See the AI-Serving project. AI-Serving is serving AI/ML models in the open standard formats PMML and ONNX with both HTTP (REST API) and gRPC endpoints.

Deploy and Manage AI/ML models at scale

See the DaaS system that deploys AI/ML models in production at scale on Kubernetes.

Support

If you have any questions about the PyPMML library, please open issues on this repository.

Feedback and contributions to the project, no matter what kind, are always very welcome.

License

PyPMML is licensed under APL 2.0.