This was made to demonstrate code on medium.

rf-explanation2.py

def data_trainer(Target_Variable,data_raw,n_valid,date_column=None):
    df_raw=data_raw
    reset_rf_samples()
    ''' This if statement is to reduce the date part'''
    if date_column:
        add_datepart(df_raw,date_column)

    train_cats(df_raw)
    df,y,nas=proc_df(df_raw,Target_Variable)
    n_trn=len(df)-n_valid
    raw_train,raw_valid=split_vals(df_raw,n_trn)
    X_train,X_valid=split_vals(df,n_trn)
    y_train,y_valid=split_vals(y,n_trn)

    ''' The Decider is the rf sampling we will be doing to speed up the process'''
    decider=None
    if len(X_train)>20000:
        decider=20000


    ''' from here we are tuning the parameters '''
    if decider:
        set_rf_samples(decider)
    score=0
    min_leaf_a=0
    max_feature_a=None
    z=None
    list1=[1,3,5,10,25]
    list2=[0.1,0.25,0.5,0.75,0.9,"sqrt","log2",1]
    for leafs in list1:
        for features in list2:
            t=auto_train(a=leafs,b=features,X_train=X_train,y_train=y_train)
            print(fx(t,X_valid=X_valid,y_valid=y_valid),leafs)
            if fx(t,X_valid=X_valid,y_valid=y_valid)>score:
                score=fx(m=t,X_valid=X_valid,y_valid=y_valid)
                min_leaf_a=leafs
                max_feature_a=features
                z=t