modssc 0.0.1

ModSSC: a modular framework for semi-supervised classification on heterogeneous data

ModSSC

ModSSC is a modular framework for semi-supervised classification across heterogeneous modalities (text, vision, tabular, graph). It is designed for academic research: clear abstractions, reproducible pipelines, and extensible method registries.

Research goals

• Provide composable "bricks" for data loading, sampling, preprocessing, graphs, and SSL methods.
• Make experiments reproducible via declarative configs and deterministic seeds.
• Support both inductive and transductive SSL with lightweight baselines and extensible APIs.

Repository map

• src/modssc: core library + CLI tools
• bench/: end-to-end benchmark runner (GitHub-only, not shipped to PyPI)
• docs/: MkDocs site (concepts, CLI, API)
• examples/, notebooks/: demos and exploratory workflows

Install

PyPI (library + CLI tools):

python -m pip install modssc

From source (recommended for benchmark runs):

git clone https://github.com/ModSSC/ModSSC
cd ModSSC
python -m pip install -e "."

Python 3.11+ is required. For modality-specific extras, see bench/README.md.

Quickstart (library + CLI)

CLI:

modssc-datasets list
modssc-sampling --help
modssc-preprocess steps list
modssc-graph build --help
modssc-inductive methods list
modssc-transductive methods list

Python:

from modssc.data_loader import load_dataset
from modssc.sampling import sample
from modssc.preprocess import preprocess

Benchmark quickstart (GitHub-only)

python -m bench.main --config bench/configs/experiments/toy_inductive.yaml
python -m bench.main --config bench/configs/experiments/toy_transductive.yaml

Artifacts land in runs/<name-timestamp>/ with:

• config.yaml (immutable config snapshot)
• run.json (metrics + metadata)
• error.txt (full traceback on failure)

Reproducibility notes

• Every run derives a timestamped run directory and a per-stage seed from a master seed.
• Preprocess and graph caches are fingerprinted by dataset + plan + seed.
• Configs are declarative YAML; they are copied into the run directory for auditability.

Citation

If you use ModSSC in research, please cite CITATION.cff.

License

MIT. See LICENSE.