alchemist-nrel 0.2.1__py3-none-any.whl → 0.3.0__py3-none-any.whl

alchemist_core/session.py

@@ -7,10 +7,14 @@ This module provides the main entry point for using ALchemist as a headless libr
 from typing import Optional, Dict, Any, List, Tuple, Callable
 import pandas as pd
 import numpy as np
+import json
+import hashlib
+from pathlib import Path
 from alchemist_core.data.search_space import SearchSpace
 from alchemist_core.data.experiment_manager import ExperimentManager
 from alchemist_core.events import EventEmitter
 from alchemist_core.config import get_logger
+from alchemist_core.audit_log import AuditLog, SessionMetadata, AuditEntry
 
 logger = get_logger(__name__)
 
@@ -48,7 +52,8 @@ class OptimizationSession:
     
     def __init__(self, search_space: Optional[SearchSpace] = None, 
                  experiment_manager: Optional[ExperimentManager] = None,
-                 event_emitter: Optional[EventEmitter] = None):
+                 event_emitter: Optional[EventEmitter] = None,
+                 session_metadata: Optional[SessionMetadata] = None):
         """
         Initialize optimization session.
         
@@ -56,11 +61,16 @@ class OptimizationSession:
             search_space: Pre-configured SearchSpace object (optional)
             experiment_manager: Pre-configured ExperimentManager (optional)
             event_emitter: EventEmitter for progress notifications (optional)
+            session_metadata: Pre-configured session metadata (optional)
         """
         self.search_space = search_space if search_space is not None else SearchSpace()
         self.experiment_manager = experiment_manager if experiment_manager is not None else ExperimentManager()
         self.events = event_emitter if event_emitter is not None else EventEmitter()
         
+        # Session metadata and audit log
+        self.metadata = session_metadata if session_metadata is not None else SessionMetadata.create()
+        self.audit_log = AuditLog()
+        
         # Link search_space to experiment_manager
         self.experiment_manager.set_search_space(self.search_space)
         
@@ -75,7 +85,7 @@ class OptimizationSession:
             'verbose': True
         }
         
-        logger.info("OptimizationSession initialized")
+        logger.info(f"OptimizationSession initialized: {self.metadata.session_id}")
     
     # ============================================================
     # Search Space Management
@@ -222,7 +232,8 @@ class OptimizationSession:
         self.events.emit('data_loaded', {'n_experiments': n_experiments, 'filepath': filepath})
     
     def add_experiment(self, inputs: Dict[str, Any], output: float, 
-                      noise: Optional[float] = None) -> None:
+                      noise: Optional[float] = None, iteration: Optional[int] = None,
+                      reason: Optional[str] = None) -> None:
         """
         Add a single experiment to the dataset.
         
@@ -230,18 +241,23 @@ class OptimizationSession:
             inputs: Dictionary mapping variable names to values
             output: Target/output value
             noise: Optional measurement uncertainty
+            iteration: Iteration number (auto-assigned if None)
+            reason: Reason for this experiment (e.g., 'Manual', 'Expected Improvement')
         
         Example:
             >>> session.add_experiment(
             ...     inputs={'temperature': 350, 'catalyst': 'A'},
-            ...     output=0.85
+            ...     output=0.85,
+            ...     reason='Manual'
             ... )
         """
         # Use ExperimentManager's add_experiment method
         self.experiment_manager.add_experiment(
             point_dict=inputs,
             output_value=output,
-            noise_value=noise
+            noise_value=noise,
+            iteration=iteration,
+            reason=reason
         )
         
         logger.info(f"Added experiment: {inputs} → {output}")
@@ -272,6 +288,88 @@ class OptimizationSession:
             'feature_names': list(X.columns)
         }
     
+    def generate_initial_design(
+        self,
+        method: str = "lhs",
+        n_points: int = 10,
+        random_seed: Optional[int] = None,
+        **kwargs
+    ) -> List[Dict[str, Any]]:
+        """
+        Generate initial experimental design (Design of Experiments).
+        
+        Creates a set of experimental conditions to evaluate before starting
+        Bayesian optimization. This does NOT add the experiments to the session -
+        you must evaluate them and add the results using add_experiment().
+        
+        Supported methods:
+        - 'random': Uniform random sampling
+        - 'lhs': Latin Hypercube Sampling (recommended, good space-filling properties)
+        - 'sobol': Sobol quasi-random sequences (low discrepancy)
+        - 'halton': Halton sequences
+        - 'hammersly': Hammersly sequences (low discrepancy)
+        
+        Args:
+            method: Sampling strategy to use
+            n_points: Number of points to generate
+            random_seed: Random seed for reproducibility
+            **kwargs: Additional method-specific parameters:
+                - lhs_criterion: For LHS method ("maximin", "correlation", "ratio")
+        
+        Returns:
+            List of dictionaries with variable names and values (no outputs)
+        
+        Example:
+            >>> # Generate initial design
+            >>> points = session.generate_initial_design('lhs', n_points=10)
+            >>> 
+            >>> # Run experiments and add results
+            >>> for point in points:
+            >>>     output = run_experiment(**point)  # Your experiment function
+            >>>     session.add_experiment(point, output=output)
+            >>> 
+            >>> # Now ready to train model
+            >>> session.train_model()
+        """
+        if len(self.search_space.variables) == 0:
+            raise ValueError(
+                "No variables defined in search space. "
+                "Use add_variable() to define variables before generating initial design."
+            )
+        
+        from alchemist_core.utils.doe import generate_initial_design
+        
+        points = generate_initial_design(
+            search_space=self.search_space,
+            method=method,
+            n_points=n_points,
+            random_seed=random_seed,
+            **kwargs
+        )
+        
+        # Store sampler info in config for audit trail
+        self.config['initial_design_method'] = method
+        self.config['initial_design_n_points'] = len(points)
+        
+        logger.info(f"Generated {len(points)} initial design points using {method} method")
+        self.events.emit('initial_design_generated', {
+            'method': method,
+            'n_points': len(points)
+        })
+        
+        # Add a lightweight audit data_locked entry for the initial design metadata
+        try:
+            extra = {'initial_design_method': method, 'initial_design_n_points': len(points)}
+            # Create an empty dataframe snapshot of the planned points
+            import pandas as pd
+            planned_df = pd.DataFrame(points)
+            self.audit_log.lock_data(planned_df, notes=f"Initial design ({method})", extra_parameters=extra)
+        except Exception:
+            # Audit logging should not block design generation
+            logger.debug("Failed to add initial design to audit log")
+
+        return points
+    
     # ============================================================
     # Model Training
     # ============================================================
@@ -437,8 +535,45 @@ class OptimizationSession:
                 # Convert complex objects to their string representation
                 json_hyperparams[key] = str(value)
         
+        # Extract kernel name and parameters
+        kernel_name = 'unknown'
+        if self.model_backend == 'sklearn':
+            # First try kernel_options
+            if hasattr(self.model, 'kernel_options') and 'kernel_type' in self.model.kernel_options:
+                kernel_name = self.model.kernel_options['kernel_type']
+                # Add nu parameter for Matern kernels
+                if kernel_name == 'Matern' and 'matern_nu' in self.model.kernel_options:
+                    json_hyperparams['matern_nu'] = self.model.kernel_options['matern_nu']
+            # Then try trained kernel
+            elif hasattr(self.model, 'model') and hasattr(self.model.model, 'kernel_'):
+                kernel_obj = self.model.model.kernel_
+                # Navigate through Product/Sum kernels to find base kernel
+                if hasattr(kernel_obj, 'k2'):  # Product kernel (Constant * BaseKernel)
+                    base_kernel = kernel_obj.k2
+                else:
+                    base_kernel = kernel_obj
+                
+                kernel_class = type(base_kernel).__name__
+                if 'Matern' in kernel_class:
+                    kernel_name = 'Matern'
+                    # Extract nu parameter if available
+                    if hasattr(base_kernel, 'nu'):
+                        json_hyperparams['matern_nu'] = float(base_kernel.nu)
+                elif 'RBF' in kernel_class:
+                    kernel_name = 'RBF'
+                elif 'RationalQuadratic' in kernel_class:
+                    kernel_name = 'RationalQuadratic'
+                else:
+                    kernel_name = kernel_class
+        elif self.model_backend == 'botorch':
+            if hasattr(self.model, 'cont_kernel_type'):
+                kernel_name = self.model.cont_kernel_type
+            elif 'kernel_type' in json_hyperparams:
+                kernel_name = json_hyperparams['kernel_type']
+        
         return {
             'backend': self.model_backend,
+            'kernel': kernel_name,
             'hyperparameters': json_hyperparams,
             'metrics': metrics,
             'is_trained': True
@@ -478,7 +613,8 @@ class OptimizationSession:
                 model=self.model,  # Pass the full SklearnModel wrapper, not just .model
                 acq_func=strategy.lower(),
                 maximize=(goal.lower() == 'maximize'),
-                random_state=self.config['random_state']
+                random_state=self.config['random_state'],
+                acq_func_kwargs=kwargs  # Pass xi, kappa, etc. to acquisition function
             )
             
             # Update acquisition with existing experimental data (un-encoded)
@@ -527,6 +663,13 @@ class OptimizationSession:
         logger.info(f"Suggested point: {suggestion_dict}")
         self.events.emit('acquisition_completed', {'suggestion': suggestion_dict})
         
+        # Cache suggestion info for audit log
+        self._last_acquisition_info = {
+            'strategy': strategy,
+            'goal': goal,
+            'parameters': kwargs
+        }
+        
         return result_df    # ============================================================
     # Predictions
     # ============================================================
@@ -594,10 +737,459 @@ class OptimizationSession:
         Update session configuration.
         
         Args:
-            **kwargs: Configuration parameters (random_state, verbose, etc.)
+            **kwargs: Configuration parameters to update
         
         Example:
             >>> session.set_config(random_state=123, verbose=False)
         """
         self.config.update(kwargs)
-        logger.info(f"Updated configuration: {kwargs}")
+        logger.info(f"Updated config: {kwargs}")
+    
+    # ============================================================
+    # Audit Log & Session Management
+    # ============================================================
+    
+    def lock_data(self, notes: str = "", extra_parameters: Optional[Dict[str, Any]] = None) -> AuditEntry:
+        """
+        Lock in current experimental data configuration.
+        
+        Creates an immutable audit log entry capturing the current data state.
+        This should be called when you're satisfied with your experimental dataset
+        and ready to proceed with modeling.
+        
+        Args:
+            notes: Optional user notes about this data configuration
+            
+        Returns:
+            Created AuditEntry
+            
+        Example:
+            >>> session.add_experiment({'temp': 100, 'pressure': 5}, output=85.2)
+            >>> session.lock_data(notes="Initial screening dataset")
+        """
+        # Set search space in audit log (once)
+        if self.audit_log.search_space_definition is None:
+            self.audit_log.set_search_space(self.search_space.variables)
+        
+        # Get current experimental data
+        df = self.experiment_manager.get_data()
+        
+        # Lock data in audit log
+        entry = self.audit_log.lock_data(
+            experiment_data=df,
+            notes=notes,
+            extra_parameters=extra_parameters
+        )
+        
+        self.metadata.update_modified()
+        logger.info(f"Locked data: {len(df)} experiments")
+        self.events.emit('data_locked', {'entry': entry.to_dict()})
+        
+        return entry
+    
+    def lock_model(self, notes: str = "") -> AuditEntry:
+        """
+        Lock in current trained model configuration.
+        
+        Creates an immutable audit log entry capturing the trained model state.
+        This should be called when you're satisfied with your model performance
+        and ready to use it for acquisition.
+        
+        Args:
+            notes: Optional user notes about this model
+            
+        Returns:
+            Created AuditEntry
+            
+        Raises:
+            ValueError: If no model has been trained
+            
+        Example:
+            >>> session.train_model(backend='sklearn', kernel='matern')
+            >>> session.lock_model(notes="Best cross-validation performance")
+        """
+        if self.model is None:
+            raise ValueError("No trained model available. Use train_model() first.")
+        
+        # Set search space in audit log (once)
+        if self.audit_log.search_space_definition is None:
+            self.audit_log.set_search_space(self.search_space.variables)
+        
+        # Get model info
+        model_info = self.get_model_summary()
+        
+        # Extract hyperparameters
+        hyperparameters = model_info.get('hyperparameters', {})
+        
+        # Get kernel name from model_info (which extracts it properly)
+        kernel_name = model_info.get('kernel', 'unknown')
+        
+        # Get CV metrics if available - use model_info metrics which are already populated
+        cv_metrics = model_info.get('metrics', None)
+        if cv_metrics and all(k in cv_metrics for k in ['rmse', 'r2']):
+            # Metrics already in correct format from get_model_summary
+            pass
+        elif hasattr(self.model, 'cv_cached_results') and self.model.cv_cached_results:
+            # Fallback to direct access
+            cv_metrics = {
+                'rmse': float(self.model.cv_cached_results.get('rmse', 0)),
+                'r2': float(self.model.cv_cached_results.get('r2', 0)),
+                'mae': float(self.model.cv_cached_results.get('mae', 0))
+            }
+        else:
+            cv_metrics = None
+        
+        # Get current iteration number
+        # Use the next iteration number for the model lock so model+acquisition share the same iteration
+        iteration = self.experiment_manager._current_iteration + 1
+        
+        # Include scaler information if available in hyperparameters
+        try:
+            if hasattr(self.model, 'input_transform_type'):
+                hyperparameters['input_transform_type'] = self.model.input_transform_type
+            if hasattr(self.model, 'output_transform_type'):
+                hyperparameters['output_transform_type'] = self.model.output_transform_type
+        except Exception:
+            pass
+
+        # Try to extract Matern nu for sklearn models if not already present
+        try:
+            if self.model_backend == 'sklearn' and 'matern_nu' not in hyperparameters:
+                # Try to navigate fitted kernel object for sklearn GaussianProcessRegressor
+                if hasattr(self.model, 'model') and hasattr(self.model.model, 'kernel_'):
+                    kernel_obj = self.model.model.kernel_
+                    base_kernel = getattr(kernel_obj, 'k2', kernel_obj)
+                    if hasattr(base_kernel, 'nu'):
+                        hyperparameters['matern_nu'] = float(base_kernel.nu)
+        except Exception:
+            pass
+
+        entry = self.audit_log.lock_model(
+            backend=self.model_backend,
+            kernel=kernel_name,
+            hyperparameters=hyperparameters,
+            cv_metrics=cv_metrics,
+            iteration=iteration,
+            notes=notes
+        )
+        
+        self.metadata.update_modified()
+        logger.info(f"Locked model: {self.model_backend}/{model_info.get('kernel')}, iteration {iteration}")
+        self.events.emit('model_locked', {'entry': entry.to_dict()})
+        
+        return entry
+    
+    def lock_acquisition(self, strategy: str, parameters: Dict[str, Any],
+                        suggestions: List[Dict[str, Any]], notes: str = "") -> AuditEntry:
+        """
+        Lock in acquisition function decision and suggested experiments.
+        
+        Creates an immutable audit log entry capturing the acquisition decision.
+        This should be called when you've reviewed the suggestions and are ready
+        to run the recommended experiments.
+        
+        Args:
+            strategy: Acquisition strategy name ('EI', 'PI', 'UCB', etc.)
+            parameters: Acquisition function parameters (xi, kappa, etc.)
+            suggestions: List of suggested experiment dictionaries
+            notes: Optional user notes about this decision
+            
+        Returns:
+            Created AuditEntry
+            
+        Example:
+            >>> suggestions = session.suggest_next(strategy='EI', n_suggestions=3)
+            >>> session.lock_acquisition(
+            ...     strategy='EI',
+            ...     parameters={'xi': 0.01, 'goal': 'maximize'},
+            ...     suggestions=suggestions,
+            ...     notes="Top 3 candidates for next batch"
+            ... )
+        """
+        # Set search space in audit log (once)
+        if self.audit_log.search_space_definition is None:
+            self.audit_log.set_search_space(self.search_space.variables)
+        
+        # Increment iteration counter first so this acquisition is logged as the next iteration
+        self.experiment_manager._current_iteration += 1
+        iteration = self.experiment_manager._current_iteration
+
+        entry = self.audit_log.lock_acquisition(
+            strategy=strategy,
+            parameters=parameters,
+            suggestions=suggestions,
+            iteration=iteration,
+            notes=notes
+        )
+        
+        self.metadata.update_modified()
+        logger.info(f"Locked acquisition: {strategy}, {len(suggestions)} suggestions")
+        self.events.emit('acquisition_locked', {'entry': entry.to_dict()})
+        
+        return entry
+    
+    def get_audit_log(self) -> List[Dict[str, Any]]:
+        """
+        Get complete audit log as list of dictionaries.
+        
+        Returns:
+            List of audit entry dictionaries
+        """
+        return self.audit_log.to_dict()
+    
+    def export_audit_markdown(self) -> str:
+        """
+        Export audit log as markdown for publications.
+        
+        Returns:
+            Markdown-formatted audit trail
+        """
+        # Pass session metadata to markdown exporter so user-entered metadata appears
+        try:
+            metadata_dict = self.metadata.to_dict()
+        except Exception:
+            metadata_dict = None
+
+        return self.audit_log.to_markdown(session_metadata=metadata_dict)
+    
+    def save_session(self, filepath: str):
+        """
+        Save complete session state to JSON file.
+        
+        Saves all session data including:
+        - Session metadata (name, description, tags)
+        - Search space definition
+        - Experimental data
+        - Trained model state (if available)
+        - Complete audit log
+        
+        Args:
+            filepath: Path to save session file (.json extension recommended)
+            
+        Example:
+            >>> session.save_session("~/ALchemist_Sessions/catalyst_study_nov2025.json")
+        """
+        filepath = Path(filepath)
+        
+        # Update audit log's experimental data snapshot to reflect current state
+        # This ensures the data table in the audit log markdown is always up-to-date
+        current_data = self.experiment_manager.get_data()
+        if current_data is not None and len(current_data) > 0:
+            self.audit_log.experiment_data = current_data.copy()
+        
+        # Prepare session data
+        session_data = {
+            'version': '1.0.0',
+            'metadata': self.metadata.to_dict(),
+            'audit_log': self.audit_log.to_dict(),
+            'search_space': {
+                'variables': self.search_space.variables
+            },
+            'experiments': {
+                'data': self.experiment_manager.get_data().to_dict(orient='records'),
+                'n_total': len(self.experiment_manager.df)
+            },
+            'config': self.config
+        }
+        
+        # Add model state if available
+        if self.model is not None:
+            model_info = self.get_model_summary()
+            
+            # Get kernel name from model_info which properly extracts it
+            kernel_name = model_info.get('kernel', 'unknown')
+            
+            # Extract kernel parameters if available
+            kernel_params = {}
+            if self.model_backend == 'sklearn' and hasattr(self.model, 'model'):
+                kernel_obj = self.model.model.kernel
+                # Extract kernel-specific parameters
+                if hasattr(kernel_obj, 'get_params'):
+                    kernel_params = kernel_obj.get_params()
+            elif self.model_backend == 'botorch':
+                # For BoTorch, parameters are in hyperparameters
+                hyperparams = model_info.get('hyperparameters', {})
+                if 'matern_nu' in hyperparams:
+                    kernel_params['nu'] = hyperparams['matern_nu']
+            
+            session_data['model_config'] = {
+                'backend': self.model_backend,
+                'kernel': kernel_name,
+                'kernel_params': kernel_params,
+                'hyperparameters': model_info.get('hyperparameters', {}),
+                'metrics': model_info.get('metrics', {})
+            }
+        
+        # Create directory if needed
+        filepath.parent.mkdir(parents=True, exist_ok=True)
+        
+        # Write JSON
+        with open(filepath, 'w') as f:
+            json.dump(session_data, f, indent=2, default=str)
+        
+        self.metadata.update_modified()
+        logger.info(f"Saved session to {filepath}")
+        self.events.emit('session_saved', {'filepath': str(filepath)})
+
+    def export_session_json(self) -> str:
+        """
+        Export current session state as a JSON string (no filesystem side-effects for caller).
+
+        Returns:
+            JSON string of session data
+        """
+        import tempfile
+        from pathlib import Path
+
+        with tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False) as tmp:
+            tmp_path = tmp.name
+            # Use existing save_session logic to write a complete JSON
+            self.save_session(tmp_path)
+
+        try:
+            with open(tmp_path, 'r') as f:
+                content = f.read()
+        finally:
+            Path(tmp_path).unlink(missing_ok=True)
+
+        return content
+    
+    @staticmethod
+    def load_session(filepath: str, retrain_on_load: bool = True) -> 'OptimizationSession':
+        """
+        Load session from JSON file.
+        
+        Args:
+            filepath: Path to session file
+            
+        Returns:
+            OptimizationSession with restored state
+            
+        Example:
+            >>> session = OptimizationSession.load_session("my_session.json")
+        """
+        filepath = Path(filepath)
+        
+        with open(filepath, 'r') as f:
+            session_data = json.load(f)
+        
+        # Check version compatibility
+        version = session_data.get('version', '1.0.0')
+        if not version.startswith('1.'):
+            logger.warning(f"Session file version {version} may not be fully compatible")
+        
+        # Create session
+        session = OptimizationSession()
+        
+        # Restore metadata
+        if 'metadata' in session_data:
+            session.metadata = SessionMetadata.from_dict(session_data['metadata'])
+        
+        # Restore audit log
+        if 'audit_log' in session_data:
+            session.audit_log.from_dict(session_data['audit_log'])
+        
+        # Restore search space
+        if 'search_space' in session_data:
+            for var in session_data['search_space']['variables']:
+                session.search_space.add_variable(
+                    var['name'],
+                    var['type'],
+                    **{k: v for k, v in var.items() if k not in ['name', 'type']}
+                )
+        
+        # Restore experimental data
+        if 'experiments' in session_data and session_data['experiments']['data']:
+            df = pd.DataFrame(session_data['experiments']['data'])
+            
+            # Metadata columns to exclude from inputs
+            metadata_cols = {'Output', 'Noise', 'Iteration', 'Reason'}
+            
+            # Add experiments one by one
+            for _, row in df.iterrows():
+                # Only include actual input variables, not metadata
+                inputs = {col: row[col] for col in df.columns if col not in metadata_cols}
+                output = row.get('Output')
+                noise = row.get('Noise') if pd.notna(row.get('Noise')) else None
+                iteration = row.get('Iteration') if pd.notna(row.get('Iteration')) else None
+                reason = row.get('Reason') if pd.notna(row.get('Reason')) else None
+                
+                session.add_experiment(inputs, output, noise=noise, iteration=iteration, reason=reason)
+        
+        # Restore config
+        if 'config' in session_data:
+            session.config.update(session_data['config'])
+        
+        # Auto-retrain model if configuration exists (optional)
+        if 'model_config' in session_data and retrain_on_load:
+            model_config = session_data['model_config']
+            logger.info(f"Auto-retraining model: {model_config['backend']} with {model_config.get('kernel', 'default')} kernel")
+            
+            try:
+                # Trigger model training with saved configuration
+                session.train_model(
+                    backend=model_config['backend'],
+                    kernel=model_config.get('kernel', 'Matern'),
+                    kernel_params=model_config.get('kernel_params', {})
+                )
+                logger.info("Model retrained successfully")
+                session.events.emit('model_retrained', {'backend': model_config['backend']})
+            except Exception as e:
+                logger.warning(f"Failed to retrain model: {e}")
+                session.events.emit('model_retrain_failed', {'error': str(e)})
+        
+        logger.info(f"Loaded session from {filepath}")
+        session.events.emit('session_loaded', {'filepath': str(filepath)})
+        
+        return session
+    
+    def update_metadata(self, name: Optional[str] = None, 
+                       description: Optional[str] = None,
+                       tags: Optional[List[str]] = None,
+                       author: Optional[str] = None):
+        """
+        Update session metadata.
+        
+        Args:
+            name: New session name (optional)
+            description: New description (optional)
+            tags: New tags (optional)
+            
+        Example:
+            >>> session.update_metadata(
+            ...     name="Catalyst Screening - Final",
+            ...     description="Optimized Pt/Pd ratios",
+            ...     tags=["catalyst", "platinum", "palladium", "final"]
+            ... )
+        """
+        if name is not None:
+            self.metadata.name = name
+        if description is not None:
+            self.metadata.description = description
+        if author is not None:
+            # Backwards compatible: store author if provided
+            setattr(self.metadata, 'author', author)
+        if tags is not None:
+            self.metadata.tags = tags
+        
+        self.metadata.update_modified()
+        logger.info("Updated session metadata")
+        self.events.emit('metadata_updated', self.metadata.to_dict())
+    
+    # ============================================================
+    # Legacy Configuration
+    # ============================================================
+    
+    def set_config(self, **kwargs) -> None:
+        """
+        Update session configuration.
+        
+        Args:
+            **kwargs: Configuration parameters to update
+        
+        Example:
+            >>> session.set_config(random_state=123, verbose=False)
+        """
+        self.config.update(kwargs)
+        logger.info(f"Updated config: {kwargs}")
+