Remove collector

Author: FilippoOlivo

pina/collector.py

@@ -12,7 +12,8 @@
 from torch.utils.data.distributed import DistributedSampler
 from ..label_tensor import LabelTensor
 from .dataset import PinaDatasetFactory, PinaTensorDataset
-from ..collector import Collector
+
+# from ..collector import Collector
 
 
 class DummyDataloader:
@@ -330,9 +331,10 @@ def __init__(
             self.pin_memory = pin_memory
 
         # Collect data
-        collector = Collector(problem)
-        collector.store_fixed_data()
-        collector.store_sample_domains()
+        # collector = Collector(problem)
+        # collector.store_fixed_data()
+        # collector.store_sample_domains()
+        problem.aggregate_data()
 
         # Check if the splits are correct
         self._check_slit_sizes(train_size, test_size, val_size)
@@ -361,7 +363,7 @@ def __init__(
             # raises NotImplementedError
             self.val_dataloader = super().val_dataloader
 
-        self.collector_splits = self._create_splits(collector, splits_dict)
+        self.data_splits = self._create_splits(problem.data, splits_dict)
         self.transfer_batch_to_device = self._transfer_batch_to_device
 
     def setup(self, stage=None):
@@ -376,23 +378,23 @@ def setup(self, stage=None):
         """
         if stage == "fit" or stage is None:
             self.train_dataset = PinaDatasetFactory(
-                self.collector_splits["train"],
+                self.data_splits["train"],
                 max_conditions_lengths=self.find_max_conditions_lengths(
                     "train"
                 ),
                 automatic_batching=self.automatic_batching,
             )
-            if "val" in self.collector_splits.keys():
+            if "val" in self.data_splits.keys():
                 self.val_dataset = PinaDatasetFactory(
-                    self.collector_splits["val"],
+                    self.data_splits["val"],
                     max_conditions_lengths=self.find_max_conditions_lengths(
                         "val"
                     ),
                     automatic_batching=self.automatic_batching,
                 )
         elif stage == "test":
             self.test_dataset = PinaDatasetFactory(
-                self.collector_splits["test"],
+                self.data_splits["test"],
                 max_conditions_lengths=self.find_max_conditions_lengths("test"),
                 automatic_batching=self.automatic_batching,
             )
@@ -473,7 +475,7 @@ def _apply_shuffle(condition_dict, len_data):
         for (
             condition_name,
             condition_dict,
-        ) in collector.data_collections.items():
+        ) in collector.items():
             len_data = len(condition_dict["input"])
             if self.shuffle:
                 _apply_shuffle(condition_dict, len_data)
@@ -540,7 +542,7 @@ def find_max_conditions_lengths(self, split):
         """
 
         max_conditions_lengths = {}
-        for k, v in self.collector_splits[split].items():
+        for k, v in self.data_splits[split].items():
             if self.batch_size is None:
                 max_conditions_lengths[k] = len(v["input"])
             elif self.repeat:

pina/data/data_module.py

@@ -23,14 +23,11 @@ def __init__(self):
         Initialization of the :class:`AbstractProblem` class.
         """
         self._discretised_domains = {}
-        # create collector to manage problem data
 
         # create hook conditions <-> problems
         for condition_name in self.conditions:
             self.conditions[condition_name].problem = self
 
-        self._batching_dimension = 0
-
         # Store in domains dict all the domains object directly passed to
         # ConditionInterface. Done for back compatibility with PINA <0.2
         if not hasattr(self, "domains"):
@@ -41,24 +38,7 @@ def __init__(self):
                     self.domains[cond_name] = cond.domain
                     cond.domain = cond_name
 
-    @property
-    def batching_dimension(self):
-        """
-        Get batching dimension.
-
-        :return: The batching dimension.
-        :rtype: int
-        """
-        return self._batching_dimension
-
-    @batching_dimension.setter
-    def batching_dimension(self, value):
-        """
-        Set the batching dimension.
-
-        :param int value: The batching dimension.
-        """
-        self._batching_dimension = value
+        self.data = None
 
     #  back compatibility 0.1
     @property
@@ -300,3 +280,26 @@ def add_points(self, new_points_dict):
             self.discretised_domains[k] = LabelTensor.vstack(
                 [self.discretised_domains[k], v]
             )
+
+    def aggregate_data(self):
+        """
+        Aggregate data from the problem's conditions into a single dictionary.
+        """
+        self.data = {}
+        if not self.are_all_domains_discretised:
+            raise RuntimeError(
+                "All domains must be discretised before aggregating data."
+            )
+        for condition_name in self.conditions:
+            condition = self.conditions[condition_name]
+            if hasattr(condition, "domain"):
+                samples = self.discretised_domains[condition.domain]
+
+                self.data[condition_name] = {
+                    "input": samples,
+                    "equation": condition.equation,
+                }
+            else:
+                keys = condition.__slots__
+                values = [getattr(condition, name) for name in keys]
+                self.data[condition_name] = dict(zip(keys, values))