Merge remote-tracking branch 'upstream/develop' into develop-explanat…

…ion-metrics
ispras · Oct 29, 2024 · b08ea72 · b08ea72
2 parents a62a3e1 + 4884213
commit b08ea72
Show file tree

Hide file tree

Showing 3 changed files with 39 additions and 15 deletions.
diff --git a/experiments/GNNExplainerGEOM_exp_example.py b/experiments/GNNExplainerGEOM_exp_example.py
@@ -23,13 +23,12 @@ def geom_GNNExplainer_test():
 
     gcn2 = model_configs_zoo(dataset=dataset, model_name='gcn_gcn')
 
-    gnn_model_manager_config = ModelManagerConfig(**{
-        "mask_features": [],
-        "optimizer": {
-            "_class_name": "Adam",
-            "_config_kwargs": {},
-        }
-    })
+    gnn_model_manager_config = ConfigPattern(
+            _config_class="ModelManagerConfig",
+            _config_kwargs={
+                "mask_features": []
+            }
+        )
 
     steps_epochs = 200
     gnn_model_manager = FrameworkGNNModelManager(

diff --git a/src/aux/declaration.py b/src/aux/declaration.py
@@ -183,6 +183,12 @@ def declare_model_by_config(
             GNNModelManager_hash: str,
             model_ver_ind: int,
             gnn_name: str,
+            mi_defense_hash: str,
+            evasion_defense_hash: str,
+            poison_defense_hash: str,
+            mi_attack_hash: str,
+            evasion_attack_hash: str,
+            poison_attack_hash: str,
             epochs=None,
     ):
         """
@@ -202,8 +208,14 @@ def declare_model_by_config(
         obj_info = {
             "gnn": gnn_name,
             "gnn_model_manager": GNNModelManager_hash,
-            "epochs": str(epochs),
             "model_ver_ind": str(model_ver_ind),
+            "poison_attacker": str(poison_attack_hash),
+            "poison_defender": str(poison_defense_hash),
+            "evasion_defender": str(evasion_defense_hash),
+            "mi_defender": str(mi_defense_hash),
+            "evasion_attacker": str(evasion_attack_hash),
+            "mi_attacker": str(mi_attack_hash),
+            "epochs": str(epochs),
         }
 
         path, files_paths = Declare.obj_info_to_path(previous_path=path, what_save=what_save,

diff --git a/src/models_builder/gnn_models.py b/src/models_builder/gnn_models.py
@@ -15,7 +15,8 @@
 
 from aux.configs import ModelManagerConfig, ModelModificationConfig, ModelConfig, CONFIG_CLASS_NAME
 from aux.data_info import UserCodeInfo
-from aux.utils import import_by_name, all_subclasses, FRAMEWORK_PARAMETERS_PATH, model_managers_info_by_names_list, hash_data_sha256, \
+from aux.utils import import_by_name, all_subclasses, FRAMEWORK_PARAMETERS_PATH, model_managers_info_by_names_list, \
+    hash_data_sha256, \
     TECHNICAL_PARAMETER_KEY, IMPORT_INFO_KEY, OPTIMIZERS_PARAMETERS_PATH, FUNCTIONS_PARAMETERS_PATH
 from aux.declaration import Declare
 from explainers.explainer import ProgressBar
@@ -247,7 +248,13 @@ class variables
                 model_ver_ind=kwargs.get('model_ver_ind') if 'model_ver_ind' in kwargs else
                 self.modification.model_ver_ind,
                 epochs=self.modification.epochs,
-                gnn_name=self.gnn.get_hash()
+                gnn_name=self.gnn.get_hash(),
+                mi_defense_hash=self.mi_defense_config.hash_for_config(),
+                evasion_defense_hash=self.evasion_defense_config.hash_for_config(),
+                poison_defense_hash=self.poison_defense_config.hash_for_config(),
+                mi_attack_hash=self.mi_attack_config.hash_for_config(),
+                evasion_attack_hash=self.evasion_attack_config.hash_for_config(),
+                poison_attack_hash=self.poison_attack_config.hash_for_config(),
             )
             path = model_dir_path / 'model'
         else:
@@ -556,6 +563,12 @@ def from_model_path(model_path, dataset_path, **kwargs):
             epochs=int(model_path['epochs']) if model_path['epochs'] != 'None' else None,
             model_ver_ind=int(model_path['model_ver_ind']),
             gnn_name=model_path['gnn'],
+            poison_attack_hash=model_path['poison_attacker'],
+            poison_defense_hash=model_path['poison_defender'],
+            evasion_defense_hash=model_path['evasion_defender'],
+            mi_defense_hash=model_path['mi_defender'],
+            evasion_attack_hash=model_path['evasion_attacker'],
+            mi_attack_hash=model_path['mi_attacker'],
         )
 
         gnn_mm_file = files_paths[1]
@@ -1182,7 +1195,7 @@ class ProtGNNModelManager(FrameworkGNNModelManager):
                 "_class_import_info": ["torch.optim"],
                 "_config_kwargs": {},
             },
-            #FUNCTIONS_PARAMETERS_PATH,
+            # FUNCTIONS_PARAMETERS_PATH,
             "loss_function": {
                 "_config_class": "Config",
                 "_class_name": "CrossEntropyLoss",
@@ -1201,7 +1214,7 @@ def __init__(self, gnn=None, dataset_path=None, **kwargs):
         _config_obj = getattr(self.manager_config, CONFIG_OBJ)
         self.clst = _config_obj.clst
         self.sep = _config_obj.sep
-        #lr = _config_obj.lr
+        # lr = _config_obj.lr
         self.early_stopping_marker = _config_obj.early_stopping
         self.proj_epochs = _config_obj.proj_epochs
         self.warm_epoch = _config_obj.warm_epoch
@@ -1318,7 +1331,8 @@ def before_epoch(self, gen_dataset):
         train_ind = [n for n, x in enumerate(gen_dataset.train_mask) if x]
         # Prototype projection
         if cur_step > self.proj_epochs and cur_step % self.proj_epochs == 0:
-            self.prot_layer.projection(self.gnn, gen_dataset.dataset, train_ind, gen_dataset.dataset.data, thrsh=self.prot_thrsh)
+            self.prot_layer.projection(self.gnn, gen_dataset.dataset, train_ind, gen_dataset.dataset.data,
+                                       thrsh=self.prot_thrsh)
         self.gnn.train()
         for p in self.gnn.parameters():
             p.requires_grad = True
@@ -1346,7 +1360,6 @@ def after_epoch(self, gen_dataset):
             self.early_stop_count = 0
             self.gnn.best_prots = self.prot_layer.prototype_graphs
 
-
     def early_stopping(self, train_loss, gen_dataset, metrics, steps):
         step = self.modification.epochs
         if self.is_best:
@@ -1355,4 +1368,4 @@ def early_stopping(self, train_loss, gen_dataset, metrics, steps):
             self.early_stop_count += 1
         last_projection = (step % self.proj_epochs == 0 and step + self.proj_epochs >= steps)
 
-        return self.early_stop_count >= self.early_stopping_marker or last_projection
+        return self.early_stop_count >= self.early_stopping_marker or last_projection