add h2o schema (#413)

* add `H2OGradientBoostingEstimator` schema and h2o schema boilerplate
* add `H2OGeneralizedLinearEstimator` schema
* add `H2ORandomForestEstimator` schema
* add `H2OTargetEncoderEstimator` schema
* add `H2OXGBoostEstimator` schema
* log models as artifacts
* test custom artifact logging
* fix other test

rubicon_ml/schema/logger.py

@@ -181,9 +181,15 @@ def log_with_schema(
                 if artifact == "self":
                     experiment.log_artifact(name=obj.__class__.__name__, data_object=obj)
             elif isinstance(artifact, dict):
-                data_object = _get_data_object(obj, artifact)
-                if data_object is not None:
-                    experiment.log_artifact(name=artifact["name"], data_object=data_object)
+                if "self" in artifact:
+                    logging_func_name = artifact["self"]
+                    logging_func = getattr(experiment, logging_func_name)
+                    logging_func(obj)
+                else:
+                    data_object = _get_data_object(obj, artifact)
+
+                    if data_object is not None:
+                        experiment.log_artifact(name=artifact["name"], data_object=data_object)
 
         for dataframe in self.schema_.get("dataframes", []):
             df_value = _get_df(obj, dataframe)

rubicon_ml/schema/registry.py

@@ -6,6 +6,21 @@
 import yaml
 
 RUBICON_SCHEMA_REGISTRY = {
+    "h2o__H2OGeneralizedLinearEstimator": lambda: _load_schema(
+        os.path.join("schema", "h2o__H2OGeneralizedLinearEstimator.yaml")
+    ),
+    "h2o__H2OGradientBoostingEstimator": lambda: _load_schema(
+        os.path.join("schema", "h2o__H2OGradientBoostingEstimator.yaml")
+    ),
+    "h2o__H2ORandomForestEstimator": lambda: _load_schema(
+        os.path.join("schema", "h2o__H2ORandomForestEstimator.yaml")
+    ),
+    "h2o__H2OTargetEncoderEstimator": lambda: _load_schema(
+        os.path.join("schema", "h2o__H2OTargetEncoderEstimator.yaml")
+    ),
+    "h2o__H2OXGBoostEstimator": lambda: _load_schema(
+        os.path.join("schema", "h2o__H2OXGBoostEstimator.yaml")
+    ),
     "lightgbm__LGBMModel": lambda: _load_schema(os.path.join("schema", "lightgbm__LGBMModel.yaml")),
     "lightgbm__LGBMClassifier": lambda: _load_schema(
         os.path.join("schema", "lightgbm__LGBMClassifier.yaml")

rubicon_ml/schema/schema/h2o__H2OGeneralizedLinearEstimator.yaml

@@ -0,0 +1,158 @@
+name: h2o__H2OGeneralizedLinearEstimator
+version: 1.0.0
+
+compatibility:
+  lightgbm:
+    max_version:
+    min_version: 3.44.0.1
+docs_url: https://docs.h2o.ai/h2o/latest-stable/h2o-py/docs/modeling.html#h2ogeneralizedlinearestimator
+
+artifacts:
+  - self: log_h2o_model
+parameters:
+  - name: alpha
+    value_attr: alpha
+  - name: auc_type
+    value_attr: auc_type
+  - name: balance_classes
+    value_attr: balance_classes
+  - name: beta_constraints
+    value_attr: beta_constraints
+  - name: beta_epsilon
+    value_attr: beta_epsilon
+  - name: build_null_model
+    value_attr: build_null_model
+  - name: calc_like
+    value_attr: calc_like
+  - name: class_sampling_factors
+    value_attr: class_sampling_factors
+  - name: cold_start
+    value_attr: cold_start
+  - name: compute_p_values
+    value_attr: compute_p_values
+  - name: custom_metric_func
+    value_attr: custom_metric_func
+  - name: dispersion_epsilon
+    value_attr: dispersion_epsilon
+  - name: dispersion_learning_rate
+    value_attr: dispersion_learning_rate
+  - name: dispersion_parameter_method
+    value_attr: dispersion_parameter_method
+  - name: early_stopping
+    value_attr: early_stopping
+  - name: export_checkpoints_dir
+    value_attr: export_checkpoints_dir
+  - name: family
+    value_attr: family
+  - name: fix_dispersion_parameter
+    value_attr: fix_dispersion_parameter
+  - name: fix_tweedie_variance_power
+    value_attr: fix_tweedie_variance_power
+  - name: fold_assignment
+    value_attr: fold_assignment
+  - name: fold_column
+    value_attr: fold_column
+  - name: gainslift_bins
+    value_attr: gainslift_bins
+  - name: generate_scoring_history
+    value_attr: generate_scoring_history
+  - name: generate_variable_inflation_factors
+    value_attr: generate_variable_inflation_factors
+  - name: gradient_epsilon
+    value_attr: gradient_epsilon
+  - name: HGLM
+    value_attr: HGLM
+  - name: ignore_const_cols
+    value_attr: ignore_const_cols
+  - name: ignored_columns
+    value_attr: ignored_columns
+  - name: influence
+    value_attr: influence
+  - name: init_dispersion_parameter
+    value_attr: init_dispersion_parameter
+  - name: interaction_pairs
+    value_attr: interaction_pairs
+  - name: interactions
+    value_attr: interactions
+  - name: intercept
+    value_attr: intercept
+  - name: keep_cross_validation_fold_assignment
+    value_attr: keep_cross_validation_fold_assignment
+  - name: keep_cross_validation_models
+    value_attr: keep_cross_validation_models
+  - name: keep_cross_validation_predictions
+    value_attr: keep_cross_validation_predictions
+  - name: lambda_
+    value_attr: lambda_
+  - name: lambda_min_ratio
+    value_attr: lambda_min_ratio
+  - name: lambda_search
+    value_attr: lambda_search
+  - name: link
+    value_attr: link
+  - name: max_active_predictors
+    value_attr: max_active_predictors
+  - name: max_after_balance_size
+    value_attr: max_after_balance_size
+  - name: max_confusion_matrix_size
+    value_attr: max_confusion_matrix_size
+  - name: max_iterations
+    value_attr: max_iterations
+  - name: max_iterations_dispersion
+    value_attr: max_iterations_dispersion
+  - name: max_runtime_secs
+    value_attr: max_runtime_secs
+  - name: missing_values_handling
+    value_attr: missing_values_handling
+  - name: nfolds
+    value_attr: nfolds
+  - name: nlambdas
+    value_attr: nlambdas
+  - name: non_negative
+    value_attr: non_negative
+  - name: obj_reg
+    value_attr: obj_reg
+  - name: objective_epsilon
+    value_attr: objective_epsilon
+  - name: offset_column
+    value_attr: offset_column
+  - name: prior
+    value_attr: prior
+  - name: rand_family
+    value_attr: rand_family
+  - name: rand_link
+    value_attr: rand_link
+  - name: random_columns
+    value_attr: random_columns
+  - name: remove_collinear_columns
+    value_attr: remove_collinear_columns
+  - name: response_column
+    value_attr: response_column
+  - name: score_each_iteration
+    value_attr: score_each_iteration
+  - name: score_iteration_interval
+    value_attr: score_iteration_interval
+  - name: seed
+    value_attr: seed
+  - name: solver
+    value_attr: solver
+  - name: standardize
+    value_attr: standardize
+  - name: startval
+    value_attr: startval
+  - name: stopping_metric
+    value_attr: stopping_metric
+  - name: stopping_rounds
+    value_attr: stopping_rounds
+  - name: stopping_tolerance
+    value_attr: stopping_tolerance
+  - name: theta
+    value_attr: theta
+  - name: tweedie_epsilon
+    value_attr: tweedie_epsilon
+  - name: tweedie_link_power
+    value_attr: tweedie_link_power
+  - name: tweedie_variance_power
+    value_attr: tweedie_variance_power
+  - name: weights_column
+    value_attr: weights_column

rubicon_ml/schema/schema/h2o__H2OGradientBoostingEstimator.yaml

@@ -0,0 +1,130 @@
+name: h2o__H2OGradientBoostingEstimator 
+version: 1.0.0
+
+compatibility:
+  lightgbm:
+    max_version:
+    min_version: 3.44.0.1
+docs_url: https://docs.h2o.ai/h2o/latest-stable/h2o-py/docs/modeling.html#h2ogradientboostingestimator
+
+artifacts:
+  - self: log_h2o_model
+parameters:
+  - name: auc_type
+    value_attr: auc_type
+  - name: auto_rebalance
+    value_attr: auto_rebalance
+  - name: balance_classes
+    value_attr: balance_classes
+  - name: build_tree_one_node
+    value_attr: build_tree_one_node
+  - name: calibrate_model
+    value_attr: calibrate_model
+  - name: calibration_method
+    value_attr: calibration_method
+  - name: categorical_encoding
+    value_attr: categorical_encoding
+  - name: check_constant_response
+    value_attr: check_constant_response
+  - name: class_sampling_factors
+    value_attr: class_sampling_factors
+  - name: col_sample_rate
+    value_attr: col_sample_rate
+  - name: col_sample_rate_change_per_level
+    value_attr: col_sample_rate_change_per_level
+  - name: col_sample_rate_per_tree
+    value_attr: col_sample_rate_per_tree
+  - name: custom_distribution_func
+    value_attr: custom_distribution_func
+  - name: custom_metric_func
+    value_attr: custom_metric_func
+  - name: distribution
+    value_attr: distribution
+  - name: export_checkpoints_dir
+    value_attr: export_checkpoints_dir
+  - name: fold_assignment
+    value_attr: fold_assignment
+  - name: fold_column
+    value_attr: fold_column
+  - name: gainslift_bins
+    value_attr: gainslift_bins
+  - name: histogram_type
+    value_attr: histogram_type
+  - name: huber_alpha
+    value_attr: huber_alpha
+  - name: ignore_const_cols
+    value_attr: ignore_const_cols
+  - name: ignored_columns
+    value_attr: ignored_columns
+  - name: in_training_checkpoints_dir
+    value_attr: in_training_checkpoints_dir
+  - name: in_training_checkpoints_tree_interval
+    value_attr: in_training_checkpoints_tree_interval
+  - name: interaction_constraints
+    value_attr: interaction_constraints
+  - name: keep_cross_validation_fold_assignment
+    value_attr: keep_cross_validation_fold_assignment
+  - name: keep_cross_validation_models
+    value_attr: keep_cross_validation_models
+  - name: keep_cross_validation_predictions
+    value_attr: keep_cross_validation_predictions
+  - name: learn_rate
+    value_attr: learn_rate
+  - name: learn_rate_annealing
+    value_attr: learn_rate_annealing
+  - name: max_abs_leafnode_pred
+    value_attr: max_abs_leafnode_pred
+  - name: max_after_balance_size
+    value_attr: max_after_balance_size
+  - name: max_confusion_matrix_size
+    value_attr: max_confusion_matrix_size
+  - name: max_depth
+    value_attr: max_depth
+  - name: max_runtime_secs
+    value_attr: max_runtime_secs
+  - name: min_rows
+    value_attr: min_rows
+  - name: min_split_improvement
+    value_attr: min_split_improvement
+  - name: monotone_constraints
+    value_attr: monotone_constraints
+  - name: nbins
+    value_attr: nbins
+  - name: nbins_cats
+    value_attr: nbins_cats
+  - name: nbins_top_level
+    value_attr: nbins_top_level
+  - name: nfolds
+    value_attr: nfolds
+  - name: ntrees
+    value_attr: ntrees
+  - name: offset_column
+    value_attr: offset_column
+  - name: pred_noise_bandwidth
+    value_attr: pred_noise_bandwidth
+  - name: quantile_alpha
+    value_attr: quantile_alpha
+  - name: r2_stopping
+    value_attr: r2_stopping
+  - name: response_column
+    value_attr: response_column
+  - name: sample_rate
+    value_attr: sample_rate
+  - name: sample_rate_per_class
+    value_attr: sample_rate_per_class
+  - name: score_each_iteration
+    value_attr: score_each_iteration
+  - name: score_tree_interval
+    value_attr: score_tree_interval
+  - name: seed
+    value_attr: seed
+  - name: stopping_metric
+    value_attr: stopping_metric
+  - name: stopping_rounds
+    value_attr: stopping_rounds
+  - name: stopping_tolerance
+    value_attr: stopping_tolerance
+  - name: tweedie_power
+    value_attr: tweedie_power
+  - name: weights_column
+    value_attr: weights_column