Quickstart¶
Ebonite can be used to reproduce arbitrary machine learning model in different environments.
Note
Don’t forget to install requirements for this example: pip install pandas scikit-learn flask flasgger
For instance, you can train sklearn model (code):
1 2 3 | reg = LogisticRegression()
data = pd.DataFrame([[1, 0], [0, 1]], columns=['a', 'b'])
reg.fit(data, [1, 0])
|
To use ebonite you need to create Ebonite client (code):
1 | ebnt = ebonite.Ebonite.local(clear=True)
|
Now you need to create task to push your model into (code):
1 | ebnt.create_model(reg, data, model_name='mymodel',
|
Great, now you can reproduce this model in different environment using this code (code):
1 | model = ebnt.get_model(project='my_project', task='regression_is_my_profession', model_name='mymodel')
|
And start a server that processes inference request like this (code):
1 2 | from ebonite.runtime import run_model_server
run_model_server(model)
|
Or create and start a docker container like this (code):
1 2 3 | uild docker image from model and run it
.build_and_run_instance(model, "sklearn_model_service",
runner_kwargs={'detach': False},
|
Full code can be found in examples/sklearn_model.