Defects of the deployment mechanism. I trained four times and did not deploy automatically.
Assume that my fourth model has the highest score, but its predict cannot be used.

At this time, I use deploy to deploy. There are three ways to deploy, most_recent, best_score, and rollback. most_recent will deploy the fourth model, best_score will also deploy the fourth model, and rollback will not work. So this project is an invalid project, right?

I can only delete my project, there is no other way.

-- just change xgboost 、xgboost_random_forest 、lightgbm、catboost
select from pgml.train(
project_name => 'Diabetes Regression',
task => 'regression',
relation_name => 'pgml.diabetes',
y_column_name => 'target',
algorithm=> 'ada_boost'
);

SELECT from pgml.deploy(
project_name => 'Diabetes Regression',
strategy => 'best_score'
);