Snowpark ML release notes¶

New features and updates¶

Model development updates:

• Implemented precision_score metric in SQL.

Bug fixes¶

• Fixed an issue where stack trace was being hidden by telemetry.

Model development fixes:

• Inferring model signatures no longer materializes the full dataframe in memory.

New features and updates¶

• Designated Snowpark ML Modeling API as Generally Available.

• New passthrough_col parameter in the Modeling API allows you to exclude specific columns, like index columns, during training or inference when not explicitly specifying input_cols.

Bug fixes¶

• Model development fixes:

  • confusion_matrix provided incorrect results when the row number could not be divided by the batch size.

New features and updates¶

• GridSearchCV and RandomizedSearchCV execution is now distributed on multi-node warehouses.

Bug fixes¶

• Model development fixes:

  • Columns were being excluded if their normalized names did not match the names specified in output_columns in OrdinalEncoder and LabelEncoder. Output columns no longer need to be valid Snowflake identifiers.

New features and updates¶

• None

Bug fixes¶

• Model development fixes:

  • Increased the column capacity of OrdinalEncoder.

New features and updates¶

• Add support for kneighbors.

• Support DecimalType as a data type.

Bug fixes¶

• Model development fixes:

  • Fix support for XGBoost and LightGBM models using SKLearn Grid Search and Randomized Search model selectors.

  • Fix metrics compatibility with Snowpark DataFrames that use Snowflake identifiers

New features and updates¶

• precision_score, recall_score, f1_score, fbeta_score, precision_recall_fscore_support, mean_absolute_error, mean_squared_error, and mean_absolute_percentage_error metric calculations are now distributed.

Bug fixes¶

• Model development fixes:

  • Fix UTF-8 decoding errors when using modeling modules on Windows.

  • Fix alias definitions causing SnowparkSQLUnexpectedAliasException in inference.

New features and updates¶

• Calculation of log_loss metric is now distributed.

Bug fixes¶

• Model development fixes:

  • Building images no longer fails with some Docker setups.

  • Embedding local ML library no longer fails when the library is imported by zipimport.

  • Update incorrect documentation about platform argument in the deploy function.

New features and updates¶

• None

Bug fixes¶

• Model development fixes:

  • Ordinal encoder can be used with mixed input column types.

  • Fix an issue when the sklearn default value is np.nan.

New features and updates¶

• Allow disabling telemetry.

Bug fixes¶

• Model development fixes:

  • Fix an error related to pandas.io.json.json_normalizer.

New features and updates¶

• Model development: Size of metrics result can exceed previous 8MB limit.

Bug fixes¶

• Model development fixes:

  • Fixed a bug when using simple imputer with NumPy >= 1.25.

  • Fixed a bug when inferring the type of label columns.

New features and updates¶

• Model development: Input dataframes can now be joined against data loaded from staged files.

• Model development: Added support for non-English languages.

Bug fixes¶

• None

New features and updates¶

• Model development: Added metrics:

  • d2_absolute_error_score

  • d2_pinball_score

  • explained_variance_score

  • mean_absolute_error

  • mean_absolute_percentage_error

  • mean_squared_error

Bug fixes¶

Model development: accuracy_score now works when given label column names that are lists of single values.

Behavior changes¶

• Model development: Changed Metrics APIs to follow scikit-learn metrics modules:

  • accuracy_score, confusion_matrix, precision_recall_fscore_support, and precision_score methods move to metrics.classification.

New features and updates¶

• Model development: Added metrics:

  • f1_score

  • fbeta_score

  • recall_score

  • roc_auc_score

  • roc_curve

  • log_loss

  • precision_recall_curve