Skip to content

Data_Module.md

Latest commit

 

History

History
58 lines (32 loc) · 3.33 KB

Data_Module.md

File metadata and controls

58 lines (32 loc) · 3.33 KB
Raw
graph LR
    PC9_3D_Dataset_Class["PC9_3D Dataset Class"]
    QM9_pyg_Dataset_Class["QM9_pyg Dataset Class"]
    PC9_QM9_pyg_Dataset_Class["PC9_QM9_pyg Dataset Class"]
    Data_Module["Data Module"]
    Data_Module -- "Provides Processed Graph Data" --> Model_Module
    Training_Engine_Module -- "Utilizes for Data Loading and Batching" --> Data_Module
    click Data_Module href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main/pseudo/Data_Module.md" "Details"
Loading

CodeBoardingDemoContact

Details

Analysis of the Data Module, a fundamental component for handling graph datasets for Graph Neural Networks (GNNs) in computational chemistry, adhering to 'Data Flow Architecture' and 'Modular Design' patterns.

PC9_3D Dataset Class

This class, found in pseudo/dataset/PC9_pyg.py, is responsible for loading, processing, and managing the PC9 graph dataset. It inherits from torch_geometric.data.InMemoryDataset, which is optimized for datasets that fit into memory. It handles the conversion of raw data (e.g., from a pickle file) into PyTorch Geometric Data objects, including atomic coordinates (pos), atomic numbers (z), and various target properties (e.g., HOMO, LUMO, gap, energy). It also provides methods for splitting the dataset into training, validation, and test sets.

Related Classes/Methods:

  • PC9_3D

QM9_pyg Dataset Class

Based on the file structure, it is highly probable that pseudo/dataset/QM9_pyg.py contains a similar InMemoryDataset implementation for the QM9 dataset, performing analogous data loading, processing, and management functions tailored to the QM9 dataset's specific structure and properties.

Related Classes/Methods:

PC9_QM9_pyg Dataset Class

This file (pseudo/dataset/PC9_QM9_pyg.py) likely contains an InMemoryDataset implementation that combines or handles both PC9 and QM9 datasets, potentially for joint training or specific experimental setups.

Related Classes/Methods:

Data Module [Expand]

Encapsulates all logic related to loading, preprocessing, and providing graph datasets (PC9, QM9) in a format suitable for PyTorch Geometric GNNs. It ensures data integrity, efficient access, and proper splitting for machine learning tasks.

Related Classes/Methods:

FAQ