Category: B1; Team name: HT; Dataset: MUTAG

[henrytsay]

Checklist

• My pull request has a clear and explanatory title.
• My pull request passes the Linting test.
• I added appropriate unit tests and I made sure the code passes all unit tests. (refer to comment below)
• My PR follows PEP8 guidelines. (refer to comment below)
• My code is properly documented, using numpy docs conventions, and I made sure the documentation renders properly.
• I linked to issues and PRs that are relevant to this PR.

Description

This PR implements Challenge B1 by creating a scalable data loading pipeline for large-scale inductive learning settings using an OnDiskPreProcessor class that enables memory-efficient processing of large datasets.

Key Features

OnDiskPreProcessor Implementation (topobench/data/preprocessor/on_disk_preprocessor.py):

• Inherits from torch_geometric.data.OnDiskDataset for on-disk storage
• Processes raw data samples one at a time, saving each to disk immediately as data_{i}.pt
• Supports topological lifting transformations via DataTransform integration
• Bypasses memory bottlenecks by avoiding loading entire datasets into RAM
• Maintains full compatibility with existing PreProcessor API and split loading methods

Configuration Support:

• Added use_on_disk_preprocessing: true flag to dataset configs
• Example config: MUTAG_ondisk.yaml
• Integrated with topobench/run.py for seamless config-based switching

Comprehensive Testing:

• 23 unit tests in test/data/preprocess/test_on_disk_preprocessor.py
• Tests cover: initialization, transforms, split loading, edge cases, and memory efficiency
• End-to-end integration tests in test/pipeline/test_on_disk_pipeline.py
• Validates model training with OnDiskPreProcessor using GCN on MUTAG dataset