9.16 ONNX
The oml.onnx class allows you to import your own
ONNX-format model, load it in the database, and score data using the prediction operators of
Oracle Machine Learning.
You can provide an onnx file (.onnx) along with its associated metadata to
load your ONNX-format model in the database and score using the existing data in the
database. You can also access your already existing ONNX model in the database. The
python module, onnx, accepts ONNX-format models with the following
characteristics:
- Techniques: The pretrained models must use one of the following
machine learning techniques:
- Regression
- Classification
- Clustering
- Embedding
- Data types: The model inputs must be one of the following Python
types:
- Numerical (float, int8, int16, int32, int64, uint8, uint16, uint32, uint64)
- String
- Input types: The model's inputs must be two-dimensional vectors of the shape [batch_size, n]. In this structure, the first dimension specifies the size of each batch, while the second dimension defines the size of each individual input.
For information on the oml.onnx class attributes and
methods, invoke help(oml.onnx) or see Oracle
Machine Learning for Python API Reference.
ONNX-format Model Examples
-
Load an ONNX-format model into the database:
Use the onnx module to load your ONNX-format models into the database. You need to provide the onnx file (.onnx) along with its associated metadata to build the model in OML4Py.
Note:
If the model's name is not provided when loading it to the database, a system-generated model name will be used. For example
model.load2db()from oml.algo import onnx model = onnx(onnx_file="path_to_model.onnx", mining_function='CLASSIFICATION', classification_prob_output="output", model_input={"tensor_of_dim_4":['Sepal_Length','Sepal_Width','Petal_Length','Petal_Width']}, apply_softmax=True, labels=['setosa','versicolor','virginica'], description={"description":"This model is useful.","author":"John Doe"}, default_on_null={'Sepal_Length':3.2,'Sepal_Width':2.5,'Petal_Length':46,'Petal_Width':1.8}) model.load2db("model1") -
Access an existing model:
An existing ONNX model in the database can be accessed using the onnx module. In the following example, a model named classif_auto owned by omluser is accessed.
from oml.algo import onnx model = onnx(model_name="classif_auto", model_owner="omluser")Once the model is loaded, you can view the model information by entering
model. -
Score data using an ONNX-format models:
In this example, an ONNX-format model loaded in the database is used for scoring:
model.predict(oml_iris.drop('Species'), supplemental_cols=oml_iris[:,['Sepal_Length', 'Sepal_Width', 'Species']], proba=True)
Final Example: Putting it All Together
import oml
import pandas as pd
from sklearn import datasets
# Load the iris data set and create a pandas.DataFrame for it.
iris = datasets.load_iris()
x = pd.DataFrame(iris.data,
columns = ['Sepal_Length','Sepal_Width',
'Petal_Length','Petal_Width'])
y = pd.DataFrame(list(map(lambda x:
{0: 'setosa', 1: 'versicolor',
2:'virginica'}[x], iris.target)),
columns = ['Species'])
try:
oml.drop('IRIS')
except:
pass
# Create the IRIS database table and the proxy object for the table.
oml_iris = oml.create(pd.concat([x, y], axis=1), table = 'IRIS')
model = oml.algo.onnx(onnx_file="model.onnx", mining_function='CLASSIFICATION', classification_prob_output="output",
model_input={"tensor_of_dim_4":['Sepal_Length','Sepal_Width','Petal_Length','Petal_Width']},
apply_softmax=True, labels=['setosa','versicolor','virginica'],
description={"description":"This model is useful.","author":"John Doe"},
default_on_null={'Sepal_Length':3.2,'Sepal_Width':2.5,'Petal_Length':4.6,'Petal_Width':1.8})
model.load2db("model1")
model.predict(oml_iris.drop('Species'),
supplemental_cols=oml_iris[:,['Sepal_Length', 'Sepal_Width', 'Species']], proba=True)
Sepal_Length Sepal_Width Species PREDICTION PROBABILITY
0 4.9 3.0 setosa setosa 1.000000
1 4.9 3.1 setosa setosa 1.000000
2 4.8 3.4 setosa setosa 1.000000
3 5.8 4.0 setosa setosa 1.000000
... ... ... ... ... ...
44 6.7 3.3 virginica versicolor 0.514706
45 6.7 3.0 virginica versicolor 0.514706
46 6.5 3.0 virginica versicolor 0.514706
47 5.9 3.0 virginica versicolor 0.514706
model.predict_proba(oml_iris.drop('Species'),
supplemental_cols = oml_iris[:,['Sepal_Length', 'Species']])
Sepal_Length Species PROBABILITY_OF_SETOSA \
0 4.4 setosa 1.0
1 4.4 setosa 1.0
2 4.5 setosa 1.0
3 4.8 setosa 1.0
... ... ... ...
42 6.7 virginica 0.0
43 6.9 versicolor 0.0
44 6.9 virginica 0.0
45 7.0 versicolor 0.0
PROBABILITY_OF_VERSICOLOR PROBABILITY_OF_VIRGINICA
0 0.000000 0.000000
1 0.000000 0.000000
2 0.000000 0.000000
3 0.000000 0.000000
... ... ...
42 0.514706 0.485294
43 0.514706 0.485294
44 0.514706 0.485294
45 0.514706 0.485294Related Topics