ModSSC documentation¶

Start here for a quick tour of ModSSC, its core capabilities, and the recommended next steps. If you're new, begin with Installation or Quickstart.

What is ModSSC¶

ModSSC is a modular framework for semi-supervised classification across heterogeneous modalities (audio, text, vision, tabular, graph) with research-focused abstractions and reproducible pipelines. ^[1]

Quick examples¶

The CLI examples below use the modssc entry points declared in pyproject.toml, and the Python examples call the data loader, sampling, and preprocess APIs from src/modssc/. ^{[2][3][4][5][6]}

CLI: ^[2][3]

modssc datasets list
modssc sampling --help

Python: ^[4][5][6]

from modssc.data_loader import load_dataset
from modssc.sampling import SamplingPlan, HoldoutSplitSpec, LabelingSpec, sample
from modssc.preprocess import PreprocessPlan, StepConfig, preprocess

ds = load_dataset("toy", download=True)
plan = SamplingPlan(split=HoldoutSplitSpec(test_fraction=0.0, val_fraction=0.2))
res, _ = sample(ds, plan=plan, seed=0, dataset_fingerprint=str(ds.meta["dataset_fingerprint"]))
pre_plan = PreprocessPlan(steps=(StepConfig(step_id="core.ensure_2d"), StepConfig(step_id="core.to_numpy")))
_ = preprocess(ds, pre_plan, seed=0, fit_indices=res.train_idx)

Key features¶

Dataset catalog with curated keys for tabular, text, vision, audio, and graph datasets. ^[7]
Optional provider backends for OpenML, Hugging Face datasets, TFDS, torchvision, torchaudio, and PyG. ^[8]
Deterministic sampling plans (holdout/k-fold + labeling strategies) with cached split artifacts. ^[5][9][10]
Deterministic preprocessing plans with step registry and optional pretrained encoders. ^[6][11][12]
Graph construction (kNN/epsilon/anchor) and graph-derived views (attribute, diffusion, structural). ^[13][14]
Inductive and transductive method registries with method IDs. ^[15][16]
CLI tools for datasets, sampling, preprocessing, graphs, augmentation, and evaluation. ^[2][3]
Benchmark runner with YAML experiment configs (GitHub-only, not shipped to PyPI). ^[17][2]

Quickstart links¶

Version¶

Current version is 0.0.3, sourced from src/modssc/__about__.py and referenced by the Hatch version config in pyproject.toml. ^[18][2]

Project status and support¶

Status is alpha (Development Status 3) in project metadata. ^[2]
Report issues via the GitHub tracker. ^[2]
Contributor guidance lives in the docs. ^[19]
Citation metadata is provided in CITATION.cff. ^[20]

Sources