MLflow set up

use_mlflow.py

import mlflow 
# all imports 


# PREPROCESSING FUNCTION

# input dataframe
# clean dataframe 
# set up train and test
# count vectorizer 
# fit the vectorizer

# set up a client to check experiments 
mlflow_client = MlflowClient(tracking_uri=mlflow_tracking_uri)
# set up models key 
classifier = MultinomialNB() 
model_name = "NaiveBayes" # name to display on UI
run_name = "NB_model" # name of the single run
exp_name = "FirstSentimentTest"  # name of the experiment
try:
    print("setting up experiment ")
    experiment = mlflow.create_experiment(name = exp_name)
    experiment_id = experiment.experiment_id
except:
    experiment = mlflow_client.get_experiment_by_name(exp_name)
    experiment_id = experiment.experiment_id


print("Set up mlflow tracking uri")
mlflow.set_tracking_uri(mlflow_tracking_uri)
# set up your model 
classifier = MultinomialNB() 
model_name = "NaiveBayes" # define the model name
run_name = "NB_model" # define the run name 
exp_name = "FirstSentimentTest" # set up the experiment family name 
# check if epxeriment exists
try:
    print("setting up experiment ")
    experiment = mlflow.create_experiment(name = exp_name)
    experiment_id = experiment.experiment_id
except:
    experiment = mlflow_client.get_experiment_by_name(exp_name)
    experiment_id = experiment.experiment_id
    
# run the tracker
with mlflow.start_run(experiment_id=experiment_id, run_name=run_name, nested=False,):
    # call the autolog
    mlflow.sklearn.autolog(log_models=True,log_input_examples=True,log_model_signatures=True, )
    # fit classifier 
    classifier.fit(X_train, y_train)
    # run predictions
    y_pred = classifier.predict(X_valid)
    fpr, tpr, thresholds = roc_curve(y_valid, y_pred)
    roc_auc = auc(fpr, tpr)
# end tracking   
mlflow.end_run()