HeatWave Release Notes
You can monitor the progress of MySQL HeatWave AutoML routines through Performance Schema queries made from a second MySQL Client session.
-
Review the following:
Progress tracking supports all MySQL HeatWave AutoML Routines. For each routine, progress tracking tracks each individual operation and stage within routines. Completed percentage values are also included.
Note
ML_MODEL_IMPORT only supports
progress tracking for models in ONNX format. It does not
support progress tracking for models in MySQL HeatWave AutoML format. See
Manage External ONNX
Models to learn more.
This example displays the stages for
ML_TRAIN only. Other MySQL HeatWave AutoML
routines may have different stages to review.
To track progress while training a machine learning model:
Connect to your DB system from two different MySQL Client sessions.
-
From the first MySQL Client session, train a machine learning model. See Train a Model. The following example trains a classification machine learning model.
mysql> CALL sys.ML_TRAIN('census_data.census_train', 'revenue', JSON_OBJECT('task', 'classification'), @census_model); -
While
ML_TRAINruns, use the second MySQL Client session and query the performance_schema.rpd_query_stats table, which stores data and statistics for for the last 100 completed queries. You can limit the output to the two most recent entries sorted byQUERY_IDin descending order. Optionally, use\Gto display the output in an easily readable format. The output is similar to the following:mysql> SELECT * FROM performance_schema.rpd_query_stats ORDER BY QUERY_ID DESC limit 2\G *************************** 131. row *************************** QUERY_ID: 140 STATEMENT_ID: 7881043 CONNECTION_ID: 41436 QUERY_TEXT: ML_TRAIN QEXEC_TEXT: { "status": "In Progress", "completedSteps": [ "Initialization", "Preprocessing" ], "completionPercentage": 22, "totalRunTime": null, "startTime": "04/22/2025, 15:09:23:287333 UTC", "progressItems": [ { "type": "Initialization", "status": "Completed", "lastUpdated": "04/22/2025, 15:09:28:235097 UTC", "completionPercentage": 100 }, { "type": "Preprocessing", "status": "Completed", "lastUpdated": "04/22/2025, 15:09:29:060563 UTC", "completionPercentage": 100 }, { "type": "Algorithm Selection", "status": "In Progress", "lastUpdated": "04/22/2025, 15:09:38:649341 UTC", "completionPercentage": 98 }, { "type": "Row Sampling", "status": "Not Started", "lastUpdated": "04/22/2025, 15:09:28:227493 UTC", "completionPercentage": 0 }, { "type": "Column Selection", "status": "Not Started", "lastUpdated": "04/22/2025, 15:09:28:227911 UTC", "completionPercentage": 0 }, { "type": "Hyperparameter Optimization", "status": "Not Started", "lastUpdated": "04/22/2025, 15:09:28:228325 UTC", "completionPercentage": 0 }, { "type": "Model Generation", "status": "Not Started", "lastUpdated": "04/22/2025, 15:09:28:228760 UTC", "completionPercentage": 0 }, { "type": "Prediction Explainer", "status": "Not Started", "lastUpdated": "04/22/2025, 15:09:28:229296 UTC", "completionPercentage": 0 }, { "type": "Model Explainer", "status": "Not Started", "lastUpdated": "04/22/2025, 15:09:28:229704 UTC", "completionPercentage": 0 } ], "details": { "trainingScore": null, "optimizationMetric": null, "options": { "task": "classification" }, "operation": "training", "user_name": "admin", "table_name": "census_train", "schema_name": "census_data", "model_handle": "census_data.census_train_admin_1745334557047", "target_column_name": "revenue" } } QKRN_TEXT: {} QEP_TEXT: {} INTERNAL_OFFLOAD_FAIL_TEXT: EXTERNAL_OFFLOAD_FAIL_TEXT: OFFLOAD_FAIL_QUERY_SUBSTR_START: 0 OFFLOAD_FAIL_QUERY_SUBSTR_LEN: 0 EXPORT_FILE_PREFIX: NULL *************************** 130. row *************************** QUERY_ID: 139 STATEMENT_ID: 7881035 CONNECTION_ID: 41436 QUERY_TEXT: ML_LOAD_TABLE QEXEC_TEXT: { "status": "Completed", "completedSteps": [], "completionPercentage": 100, "totalRunTime": "3.77 Seconds", "startTime": "04/22/2025, 15:09:18:046900 UTC", "progressItems": [], "details": {} } QKRN_TEXT: {} QEP_TEXT: {} INTERNAL_OFFLOAD_FAIL_TEXT: EXTERNAL_OFFLOAD_FAIL_TEXT: OFFLOAD_FAIL_QUERY_SUBSTR_START: 0 OFFLOAD_FAIL_QUERY_SUBSTR_LEN: 0 EXPORT_FILE_PREFIX: NULL 131 rows in set (0.0753 sec)Refer to the second entry to confirm that the table was successfully loaded. It shows that the
ML_LOAD_TABLEoperation has astatusofCompleted, and acompletionPercentageof100. Refer to the first entry to review the status and completion percentage for each stage ofML_TRAIN. It shows that theML_TRAINoperation has astatusofIn Progress, with sub-tasks displaying values forcompletionPercentage. -
When the status shows
Completedfor allML_TRAINtasks, run the same query again from either MySQL Client session forperformance_schema.rpd_query_stats. Limit the query to display the most recentQUERY_ID, which for this example is143. The output is similar to the following:mysql> SELECT * FROM performance_schema.rpd_query_stats WHERE QUERY_ID=143\G *************************** 134. row *************************** QUERY_ID: 143 STATEMENT_ID: 7884317 CONNECTION_ID: 41436 QUERY_TEXT: ML_UNLOAD_MODEL QEXEC_TEXT: { "status": "Completed", "completedSteps": [], "completionPercentage": 100, "totalRunTime": "0.64 Seconds", "startTime": "04/22/2025, 15:13:37:392368 UTC", "progressItems": [], "details": { "operation": "unload_model", "user_name": "admin", "model_handle": "census_data.census_train_admin_1745334557047" } } QKRN_TEXT: {} QEP_TEXT: {} INTERNAL_OFFLOAD_FAIL_TEXT: EXTERNAL_OFFLOAD_FAIL_TEXT: OFFLOAD_FAIL_QUERY_SUBSTR_START: 0 OFFLOAD_FAIL_QUERY_SUBSTR_LEN: 0 EXPORT_FILE_PREFIX: NULL 1 row in set (0.0729 sec)Confirm that the
statusshowsCompletedand thecompletionPercentageshows100.
You have the option to query only the status or completed percentage of a routine. This example loads and generates a table of predictions for the model previously trained.
Connect to your DB system from two different MySQL Client sessions.
-
From the first MySQL Client session, train a machine learning model. See Train a Model. The following example trains a classification machine learning model.
mysql> CALL sys.ML_TRAIN('census_data.census_train', 'revenue', JSON_OBJECT('task', 'classification'), @census_model); -
Load the trained model. Replace
@census_modelwith your own model handle. See Load a Model.mysql> CALL sys.ML_MODEL_LOAD(@census_model, NULL); -
Use
ML_PREDICT_TABLEto generate a table of predictions. See Generate Predictions for a Table. For example:mysql> CALL sys.ML_PREDICT_TABLE('census_data.census_train', @census_model, 'census_data.census_train_predictions', NULL); -
While the routine runs, use the second MySQL Client session and create a session variable for
QEXEC_TEXT. Query theQUERY_TEXTcolumn forML_PREDICT_TABLE, order the table byQUERY_IDin descending order, and only query the most recent entry.mysql> SELECT QEXEC_TEXT INTO @progress_json FROM performance_schema.rpd_query_stats WHERE QUERY_TEXT='ML_PREDICT_TABLE' ORDER BY QUERY_ID DESC limit 1; -
From the second MySQL Client session, query the completion percentage for the routine.
mysql> SELECT JSON_EXTRACT(@progress_json, '$.completionPercentage'); +--------------------------------------------------------+ | JSON_EXTRACT(@progress_json, '$.completionPercentage') | +--------------------------------------------------------+ | 67 | +--------------------------------------------------------+ 1 row in set (0.0463 sec) -
Query the current status for the routine.
mysql> SELECT JSON_EXTRACT(@progress_json, '$.status'); +------------------------------------------+ | JSON_EXTRACT(@progress_json, '$.status') | +------------------------------------------+ | "In Progress" | +------------------------------------------+ 1 row in set (0.0439 sec)
Review Machine Learning Use Cases to create machine learning models with sample datasets.