pgml.deploy()
Deployments
A model is automatically deployed and used for predictions if its key metric (R2 for regression, F1 for classification) is improved during training over the previous version. Alternatively, if you want to manage deploys manually, you can always change which model is currently responsible for making predictions.
API
pgml.deploy(
project_name TEXT,
strategy TEXT DEFAULT 'best_score',
algorithm TEXT DEFAULT NULL
)
Parameters
| Parameter | Example | Description |
|---|---|---|
project_name |
'My First PostgresML Project' |
The name of the project used in pgml.train() and pgml.predict(). |
strategy |
'rollback' |
The deployment strategy to use for this deployment. |
algorithm |
'xgboost' |
Restrict the deployment to a specific algorithm. Useful when training on multiple algorithms and hyperparameters at the same time. |
Strategies
There are 3 different deployment strategies available:
| Strategy | Description |
|---|---|
most_recent |
The most recently trained model for this project is immediately deployed, regardless of metrics. |
best_score |
The model that achieved the best key metric score is immediately deployed. |
rollback |
The model that was deployed before to the current one is deployed. |
The default deployment behavior allows any algorithm to qualify. It's automatically used during training, but can be manually executed as well:
Examples
Deploying the best score
SQL
SELECT * FROM pgml.deploy(
'Handwritten Digit Image Classifier',
strategy => 'best_score'
);
Output
project | strategy | algorithm
------------------------------------+------------+-----------
Handwritten Digit Image Classifier | best_score | xgboost
(1 row)
Specific Algorithms
Deployment candidates can be restricted to a specific algorithm by including the algorithm parameter. This is useful when you're training multiple algorithms using different hyperparameters and want to restrict the deployment a single algorithm only:
SQL
SELECT * FROM pgml.deploy(
project_name => 'Handwritten Digit Image Classifier',
strategy => 'best_score',
algorithm => 'svm'
);
Output
project_name | strategy | algorithm
------------------------------------+----------------+----------------
Handwritten Digit Image Classifier | classification | svm
(1 row)
Rolling Back
In case the new model isn't performing well in production, it's easy to rollback to the previous version. A rollback creates a new deployment for the old model. Multiple rollbacks in a row will oscillate between the two most recently deployed models, making rollbacks a safe and reversible operation.
Rollback
SELECT * FROM pgml.deploy(
'Handwritten Digit Image Classifier',
strategy => 'rollback'
);
Output
project | strategy | algorithm
------------------------------------+----------+-----------
Handwritten Digit Image Classifier | rollback | linear
(1 row)
Rollback again
Rollbacks are actually new deployments, so issuing two rollbacks in a row will leave you back with the original model, making rollback safely undoable.
SELECT * FROM pgml.deploy(
'Handwritten Digit Image Classifier',
strategy => 'rollback'
);
Output
project | strategy | algorithm
------------------------------------+----------+-----------
Handwritten Digit Image Classifier | rollback | xgboost
(1 row)
Specific Model IDs
In the case you need to deploy an exact model that is not the most_recent or best_score, you may deploy a model by id. Model id's can be found in the pgml.models table.
SQL
SELECT * FROM pgml.deploy(12);
Output
project | strategy | algorithm
------------------------------------+----------+-----------
Handwritten Digit Image Classifier | specific | xgboost
(1 row)
