dump/load IsolationForest to/from ONNX

dump.py

#!/usr/bin/env python3

import numpy as np

from sklearn.ensemble import IsolationForest
# Need skl2onnx 1.7.1+
from skl2onnx import convert_sklearn
from skl2onnx import to_onnx

rng = np.random.default_rng()
x1 = rng.multivariate_normal(np.array([10,10]), np.diag([1,1]), 20)
x2 = rng.multivariate_normal(np.array([-10,-10]), np.diag([1,1]), 20)

X = np.vstack([x1,x2])
X = X.astype(np.float32)

clf = IsolationForest(random_state=0).fit(X)

model_onnx = to_onnx(clf, X)

with open("isof.onnx", "wb") as f:
    f.write(model_onnx.SerializeToString())

load.py

#!/usr/bin/env python

import onnxruntime as rt
import numpy as np

X_test = np.array([[0.0, 0.0], [10.0, 10.0], [-10.0, -10.0], [10.0, 0.0]])

sess = rt.InferenceSession("isof.onnx")

print("Inputs: {}".format([x.name for x in sess.get_inputs()]))
print("Outputs: {}".format([x.name for x in sess.get_outputs()]))

input_name = sess.get_inputs()[0].name
label_name = 'scores' 

pred_onx = sess.run([label_name], {input_name: X_test.astype(np.float32)})

print(pred_onx)