How To Save & Reload a Python Machine Learning Model using Pickle ?
In this post lets understand How To Save a Machine Learning Model in Python using Pickle ? Pickle is used quite often in Python especially in Machine Learning modelling. We will try to understand what is & How to use Pickle.
What is Pickle
- Pickle is a Standard way of serializing objects in Python. It can be a Machine Learning Algorithm or any other Object.
- You can serialize and save the model or Object using Pickle . It is saved in a serialized format as a file.
- When you need to re-use or re-load the same Model or Object , you can reload and de-serialize the file using Pickle.
Example Code
- Consider a sample Logistic Regression Machine Learning model . It is taken from Scikit-learn official documentation (https://scikit-learn.org/stable/tutorial/statistical_inference/supervised_learning.html)
#======= MODEL BUILDING
import numpy as np
from sklearn import datasets
iris \= datasets.load\_iris()
iris\_X \= iris.data
iris\_y \= iris.target
np.unique(iris\_y)
\# Split iris data in train and test data
\# A random permutation, to split the data randomly
np.random.seed(0)
indices \= np.random.permutation(len(iris\_X))
iris\_X\_train \= iris\_X\[indices\[:\-10\]\]
iris\_y\_train \= iris\_y\[indices\[:\-10\]\]
iris\_X\_test \= iris\_X\[indices\[\-10:\]\]
iris\_y\_test \= iris\_y\[indices\[\-10:\]\]
\# Create and fit a nearest-neighbor classifier
from sklearn.neighbors import KNeighborsClassifier
knn\_model \= KNeighborsClassifier()
knn\_model.fit(iris\_X\_train, iris\_y\_train)
Create a Pickle File
- Our Objecctive is to create a Pickle file of the TRAINED model - knn_model in this case**.** That is we will save the model as a serialized object using Pickle.
- Use the code below -
\# save the knn\_model to disk
filename = 'Our\_Trained\_knn\_model.sav'
pickle.dump(knn\_model, open(filename, 'wb'))
- We have saved our Trained model knn_model as file Our_Trained_knn_model.sav
Re-load the Pickle File
- Our objective is to re-load the Pickle file from disk, De-serialize it and convert it back to our Knn Model to do prediction .
- All these objectives can be achieved with Pickle
- Code is below -
\# load the model from disk
filename = 'Our\_Trained\_knn\_model.sav'
knn\_model\_reloaded = pickle.load(open(filename, 'rb'))
knn\_model\_reloaded.predict(iris\_X\_test)
print(iris\_y\_test)
Other Interesting Reads -
How to log an error in Python ?
How to Code Custom Exception Handling in Python ?
How to Handle Errors and Exceptions in Python ?
Fix Various Indentation Errors in Python