Saving sklearn example

.gitignore

@@ -138,3 +138,4 @@ dmypy.json
 # Cython debug symbols
 cython_debug/
 
+./*/output

dev-requirements.in

@@ -0,0 +1,3 @@
+black
+flake8
+pip-tools

dev-requirements.txt

@@ -0,0 +1,40 @@
+#
+# This file is autogenerated by pip-compile with python 3.8
+# To update, run:
+#
+#    pip-compile dev-requirements.in
+#
+black==22.1.0
+    # via -r dev-requirements.in
+click==8.0.4
+    # via
+    #   black
+    #   pip-tools
+flake8==4.0.1
+    # via -r dev-requirements.in
+mccabe==0.6.1
+    # via flake8
+mypy-extensions==0.4.3
+    # via black
+pathspec==0.9.0
+    # via black
+pep517==0.12.0
+    # via pip-tools
+pip-tools==6.5.1
+    # via -r dev-requirements.in
+platformdirs==2.5.1
+    # via black
+pycodestyle==2.8.0
+    # via flake8
+pyflakes==2.4.0
+    # via flake8
+tomli==2.0.1
+    # via black
+typing-extensions==4.1.1
+    # via black
+wheel==0.37.1
+    # via pip-tools
+
+# The following packages are considered to be unsafe in a requirements file:
+# pip
+# setuptools

requirements.in

@@ -0,0 +1,7 @@
+joblib
+numpy
+onnx
+onnxruntime
+opencv-python
+skl2onnx
+sklearn

requirements.txt

@@ -0,0 +1,55 @@
+#
+# This file is autogenerated by pip-compile with python 3.8
+# To update, run:
+#
+#    pip-compile requirements.in
+#
+flatbuffers==2.0
+    # via onnxruntime
+joblib==1.1.0
+    # via
+    #   -r requirements.in
+    #   scikit-learn
+numpy==1.22.3
+    # via
+    #   -r requirements.in
+    #   onnx
+    #   onnxconverter-common
+    #   onnxruntime
+    #   opencv-python
+    #   scikit-learn
+    #   scipy
+    #   skl2onnx
+onnx==1.11.0
+    # via
+    #   -r requirements.in
+    #   onnxconverter-common
+    #   skl2onnx
+onnxconverter-common==1.9.0
+    # via skl2onnx
+onnxruntime==1.10.0
+    # via -r requirements.in
+opencv-python==4.5.5.64
+    # via -r requirements.in
+protobuf==3.19.4
+    # via
+    #   onnx
+    #   onnxconverter-common
+    #   onnxruntime
+    #   skl2onnx
+scikit-learn==1.0.2
+    # via
+    #   skl2onnx
+    #   sklearn
+scipy==1.8.0
+    # via
+    #   scikit-learn
+    #   skl2onnx
+skl2onnx==1.11
+    # via -r requirements.in
+sklearn==0.0
+    # via -r requirements.in
+threadpoolctl==3.1.0
+    # via scikit-learn
+typing-extensions==4.1.1
+    # via onnx

scikit_learn_ex/output/.gitignore

@@ -0,0 +1,2 @@
+*.pkl
+*.onnx

scikit_learn_ex/train.py

@@ -0,0 +1,22 @@
+from turtle import pd
+import joblib as jl
+
+from sklearn.datasets import load_iris
+from sklearn.model_selection import train_test_split
+from sklearn.ensemble import RandomForestClassifier
+
+def main():
+    # Use iris dataset
+    iris = load_iris()
+    X, y = iris.data, iris.target
+    X_train, X_test, y_train, y_test = train_test_split(X, y)
+    clr = RandomForestClassifier()
+    # Fit
+    clr.fit(X_train, y_train)
+    # Serialize the classifier to pickle file
+    jl.dump(clr, "./output/model.pkl", compress=9)
+
+if __name__ == "__main__":
+    print("Building iris model...")
+    main()
+    print("Model trained and dumped as pickle file.")

sklearn_ex/README.md

@@ -0,0 +1,27 @@
+# sklearn example
+
+This is a simple demo showing the training of Randmon Forest Classifier, serializing it as a pickle file, converting it to an Onnx model, then making inferences using the Onnx model.
+
+## Usage
+
+Train the model
+
+```
+python train.py
+```
+
+Convert to Onnx
+
+```
+
+```
+
+Make inferences
+
+```
+
+```
+
+## Model
+
+![Model Graph](./img/model_graph.png)

sklearn_ex/convert_to_onnx.py

@@ -0,0 +1,48 @@
+import argparse
+import pathlib
+
+import joblib as jl
+
+from skl2onnx import convert_sklearn
+from skl2onnx.common.data_types import FloatTensorType
+
+
+def convert_model_to_onnx(path):
+    """Converts the example sklearn classifier to Onnx"""
+    clr_model = jl.load(path)
+    # Define Initial Types
+    initial_type = [("float_input", FloatTensorType([None, 4]))]
+    onnx_model = convert_sklearn(clr_model, initial_types=initial_type)
+    return onnx_model
+
+
+def write_onnx_file(model, output_path):
+    """Writes an Onnx model to a desired file location"""
+    with open(output_path, mode="wb") as f:
+        f.write(model.SerializeToString())
+
+
+def main(args):
+    print(f"Converting {args.model_path} to Onnx")
+    model = convert_model_to_onnx(args.model_path)
+    print(f"Writing saving Onnx file to {args.output_path}")
+    write_onnx_file(model, args.output_path)
+    print("Done")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Converts a pickled classifier to onnx format"
+    )
+    parser.add_argument(
+        "model_path", type=pathlib.Path, help="Path to classifier pickle file"
+    )
+    parser.add_argument(
+        "-o" "--out",
+        type=pathlib.Path,
+        dest="output_path",
+        default=pathlib.Path("./output/clr_model.onnx"),
+        help="Path to output Onnx file",
+    )
+    args = parser.parse_args()
+    main(args)

sklearn_ex/output/.gitignore

@@ -0,0 +1,2 @@
+*.pkl
+*.onnx

sklearn_ex/train.py

@@ -0,0 +1,24 @@
+from turtle import pd
+import joblib as jl
+
+from sklearn.datasets import load_iris
+from sklearn.model_selection import train_test_split
+from sklearn.ensemble import RandomForestClassifier
+
+
+def main():
+    # Use iris dataset
+    iris = load_iris()
+    X, y = iris.data, iris.target
+    X_train, X_test, y_train, y_test = train_test_split(X, y)
+    clr = RandomForestClassifier()
+    # Fit
+    clr.fit(X_train, y_train)
+    # Serialize the classifier to pickle file
+    jl.dump(clr, "./output/model.pkl", compress=9)
+
+
+if __name__ == "__main__":
+    print("Building iris model...")
+    main()
+    print("Model trained and dumped as pickle file.")