hpsearch fixes

xsk07 · Sep 7, 2024 · 6f7de5c · 6f7de5c
1 parent 74ddd1f
commit 6f7de5c
Show file tree

Hide file tree

Showing 2 changed files with 40 additions and 20 deletions.
diff --git a/hp_search.py b/hp_search.py
@@ -1,7 +1,11 @@
 import sys
 
 from src.dataset.dataset_factory import DatasetFactory
+from src.evaluation.evaluation_metric_factory import EvaluationMetricFactory
+from src.explainer.explainer_factory import ExplainerFactory
+from src.oracle.embedder_factory import EmbedderFactory
 from src.oracle.oracle_factory import OracleFactory
+from src.plotters.plotter_factory import PlotterFactory
 from src.utils.context import Context
 from hpsearch.hp_tuner import HpTuner
 
@@ -16,11 +20,11 @@
 
     context.factories['datasets'] = DatasetFactory(context)
     context.factories['oracles'] = OracleFactory(context)
-    #context.factories['embedders'] = EmbedderFactory(context)
-    #context.factories['explainers'] = ExplainerFactory(context)
-    #context.factories['metrics'] = EvaluationMetricFactory(context.conf)
-    #context.factories['plotters'] = PlotterFactory(context)
+    context.factories['embedders'] = EmbedderFactory(context)
+    context.factories['explainers'] = ExplainerFactory(context)
+    context.factories['metrics'] = EvaluationMetricFactory(context.conf)
+    context.factories['plotters'] = PlotterFactory(context)
 
-    hp_tuner = HpTuner(context, 100)
+    hp_tuner = HpTuner(context, train_oracle=True, train_expainer=False, n_trials=100)
     hp_tuner.optimize()
 
diff --git a/hpsearch/hp_tuner.py b/hpsearch/hp_tuner.py
@@ -3,7 +3,7 @@
 
 class HpTuner():
 
-    def __init__(self, context: Context, n_trials: int) -> None:
+    def __init__(self, context: Context, train_oracle: bool, train_expainer: bool, n_trials: int) -> None:
         self.context = context
         self.n_trials = n_trials
         self.dataset = context.factories['datasets'].get_dataset(context.conf['do-pairs'][0]['dataset'])
@@ -20,6 +20,17 @@ def optimize(self):
 
         self.logger.info("Start optimization.")
 
+        #search_space = {
+        #    'num_conv_layers': [1, 2, 3, 4, 5],
+        #    'num_dense_layers': [1, 2, 3, 4, 5],
+        #    'conv_booster': [1, 2, 3, 4, 5]
+        #}
+
+        # Define the GridSampler with the search space
+        #sampler = optuna.samplers.GridSampler(search_space)
+
+        # Create a study object and use grid search
+        #study = optuna.create_study(sampler=sampler, direction='maximize')
 
         #Create a study object and optimize the objective function
         study = optuna.create_study(direction='maximize')
@@ -37,26 +48,31 @@ def objective(self, trial):
 
         #Suggest values of the hyperparameters using a trial object.
         #batch_size = trial.trial.suggest_categorical('batch_size', [32, 64])
-        learning_rate = trial.suggest_float('lr', 1e-3, 1e-1)
-        weight_decay = trial.suggest_float('weight_decay', 1e-5, 1e-1)
-        num_conv_layers =  trial.suggest_int('num_conv_layers', 1, 2)
-        num_dense_layers = trial.suggest_int('num_dense_layers', 1, 2)
-        conv_booster = trial.suggest_int('conv_booster', 1, 2)
+        learning_rate = trial.suggest_float('lr', 1e-5, 1, log=True)
+        weight_decay = trial.suggest_float('weight_decay', 1e-5, 1, log=True)
+        #num_conv_layers =  trial.suggest_int('num_conv_layers', 1, 5)
+        #num_dense_layers = trial.suggest_int('num_dense_layers', 1, 5)
+        #conv_booster = trial.suggest_int('conv_booster', 1, 5)
         linear_decay = trial.suggest_float('linear_decay', 0, 2)
 
         self.oracle_config['parameters']['optimizer']['parameters']['lr'] = learning_rate
         self.oracle_config['parameters']['optimizer']['parameters']['weight_decay'] = weight_decay
-        self.oracle_config['parameters']['model']['parameters']['num_conv_layers'] = num_conv_layers
-        self.oracle_config['parameters']['model']['parameters']['num_dense_layers'] = num_dense_layers
-        self.oracle_config['parameters']['model']['parameters']['conv_booster'] = conv_booster
+        self.oracle_config['parameters']['model']['parameters']['num_conv_layers'] = 4 #num_conv_layers
+        self.oracle_config['parameters']['model']['parameters']['num_dense_layers'] = 2 #num_dense_layers
+        self.oracle_config['parameters']['model']['parameters']['conv_booster'] = 5 #conv_booster
         self.oracle_config['parameters']['model']['parameters']['linear_decay'] = linear_decay
 
         dataset = self.context.factories['datasets'].get_dataset(self.dataset_config)
         parameters = self.oracle_config['parameters']
-        self.logger.info(f'Parameters: {parameters}')
+        self.logger.info(f'Trial {trial.number}: Hyperparameters: {parameters}')
         oracle = self.context.factories['oracles'].get_oracle(self.oracle_config, dataset)
-
-        mean_accuracy = oracle.mean_accuracy
-        if mean_accuracy is None:
-            raise ValueError("The oracle did not return a valid accuracy value.")
-        return mean_accuracy
+
+        # Check for accuracy metric
+        if hasattr(oracle, 'mean_accuracy'):
+            mean_accuracy = oracle.mean_accuracy
+            if mean_accuracy is None:
+                self.logger.error("The oracle did not return a valid accuracy value.")
+            else:
+                return mean_accuracy
+        else:
+            self.logger.error("Oracle does not have 'mean_accuracy' attribute.")