ADD: Dataset OGBG molecules

configs/dataset/graph/MOLHIV.yaml

@@ -0,0 +1,35 @@
+
+loader:
+  _target_: topobenchmark.data.loaders.OGBGDatasetLoader
+  parameters: 
+    data_domain: graph
+    data_type: OGBGDataset
+    data_name: MOLHIV
+    data_dir: ${paths.data_dir}/${dataset.loader.parameters.data_domain}/${dataset.loader.parameters.data_type}
+
+# Dataset parameters
+parameters:
+  num_features: 9
+  num_classes: 2
+  task: classification
+  loss_type: cross_entropy
+  monitor_metric: accuracy
+  task_level: graph
+
+#splits
+split_params:
+  learning_setting: inductive
+  data_split_dir: ${paths.data_dir}/data_splits/${dataset.loader.parameters.data_name}
+  data_seed: 0
+  split_type: random #'k-fold' # either "k-fold" or "random" strategies
+  k: 10 # for "k-fold" Cross-Validation
+  train_prop: 0.5 # for "random" strategy splitting
+
+# Dataloader parameters
+dataloader_params:
+  batch_size: 64 
+  num_workers: 0
+  pin_memory: False
+
+
+

configs/dataset/graph/MOLPCBA.yaml

@@ -0,0 +1,35 @@
+
+loader:
+  _target_: topobenchmark.data.loaders.OGBGDatasetLoader
+  parameters: 
+    data_domain: graph
+    data_type: OGBGDataset
+    data_name: MOLPCBA
+    data_dir: ${paths.data_dir}/${dataset.loader.parameters.data_domain}/${dataset.loader.parameters.data_type}
+
+# Dataset parameters
+parameters:
+  num_features: 9
+  num_classes: 2
+  task: classification
+  loss_type: cross_entropy
+  monitor_metric: accuracy
+  task_level: graph
+
+#splits
+split_params:
+  learning_setting: inductive
+  data_split_dir: ${paths.data_dir}/data_splits/${dataset.loader.parameters.data_name}
+  data_seed: 0
+  split_type: random #'k-fold' # either "k-fold" or "random" strategies
+  k: 10 # for "k-fold" Cross-Validation
+  train_prop: 0.5 # for "random" strategy splitting
+
+# Dataloader parameters
+dataloader_params:
+  batch_size: 64 
+  num_workers: 0
+  pin_memory: False
+
+
+

configs/dataset/graph/PPA.yaml

@@ -0,0 +1,35 @@
+
+loader:
+  _target_: topobenchmark.data.loaders.OGBGDatasetLoader
+  parameters: 
+    data_domain: graph
+    data_type: OGBGDataset
+    data_name: PPA
+    data_dir: ${paths.data_dir}/${dataset.loader.parameters.data_domain}/${dataset.loader.parameters.data_type}
+
+# Dataset parameters
+parameters:
+  num_features: 37
+  num_classes: 37
+  task: classification
+  loss_type: cross_entropy
+  monitor_metric: accuracy
+  task_level: graph
+
+#splits
+split_params:
+  learning_setting: inductive
+  data_split_dir: ${paths.data_dir}/data_splits/${dataset.loader.parameters.data_name}
+  data_seed: 0
+  split_type: random #'k-fold' # either "k-fold" or "random" strategies
+  k: 10 # for "k-fold" Cross-Validation
+  train_prop: 0.5 # for "random" strategy splitting
+
+# Dataloader parameters
+dataloader_params:
+  batch_size: 64 
+  num_workers: 0
+  pin_memory: False
+
+
+

topobenchmark/data/loaders/graph/ogbg_datasets.py

@@ -0,0 +1,85 @@
+"""Loaders for  Graph Property Prediction datasets."""
+
+import os
+from pathlib import Path
+
+from ogb.graphproppred import PygGraphPropPredDataset
+from omegaconf import DictConfig
+from torch_geometric.data import Dataset
+
+from topobenchmark.data.loaders.base import AbstractLoader
+
+
+class OGBGDatasetLoader(AbstractLoader):
+    """Load molecule datasets (molhiv, molpcba, ppa) with predefined splits.
+
+    Parameters
+    ----------
+    parameters : DictConfig
+        Configuration parameters containing:
+            - data_dir: Root directory for data
+            - data_name: Name of the dataset
+            - data_type: Type of the dataset (e.g., "molecule")
+    """
+
+    def __init__(self, parameters: DictConfig) -> None:
+        super().__init__(parameters)
+        self.datasets: list[Dataset] = []
+
+    def load_dataset(self) -> Dataset:
+        """Load the molecule dataset with predefined splits.
+
+        Returns
+        -------
+        Dataset
+            The combined dataset with predefined splits.
+
+        Raises
+        ------
+        RuntimeError
+            If dataset loading fails.
+        """
+
+        split_idx = self._load_splits()
+        combined_dataset = self._combine_splits()
+        combined_dataset.split_idx = split_idx
+        return combined_dataset
+
+    def _load_splits(self) -> None:
+        """Load the dataset splits for the specified dataset.
+
+        Returns
+        -------
+        dict
+            The split indices for the dataset.
+        """
+        dataset = PygGraphPropPredDataset(
+            name="ogbg-" + self.parameters.data_name.lower()
+        )
+        split_idx = dataset.get_idx_split()
+
+        for split in ["train", "valid", "test"]:
+            ds = dataset[split_idx[split]]
+            ds.x = ds.x.long()
+            self.datasets.append(ds)
+        return split_idx
+
+    def _combine_splits(self) -> Dataset:
+        """Combine the dataset splits into a single dataset.
+
+        Returns
+        -------
+        Dataset
+            The combined dataset containing all splits.
+        """
+        return self.datasets[0] + self.datasets[1] + self.datasets[2]
+
+    def get_data_dir(self) -> Path:
+        """Get the data directory.
+
+        Returns
+        -------
+        Path
+            The path to the dataset directory.
+        """
+        return os.path.join(self.root_data_dir, self.parameters.data_name)