v0.1.0

Added

• FairModel — PyTorch nn.Module with differentiable fairness layer (cvxpylayers) enforcing hard per-batch constraints during training
• FairModel.wrap() — wrap any existing nn.Module with fairness constraints; prediction bounds inferred automatically via dry-run forward pass
• FairTrainer — high-level training loop with early stopping, LR scheduling, checkpoint save/load, and per-epoch fairness gap tracking
• Three built-in fairness metrics: MeanPredictionParity ('mean_pred'), MeanResidualFairness ('mean_residual'), EqualizedOdds ('equalized_odds')
• FairnessMetric abstract base class for custom affine fairness constraints
• validate_metric() — pre-training DPP compliance check for custom metrics
• create_stratified_dataloaders() — stratified batching that maintains constant group ratios per batch, accepting both numpy arrays and torch tensors
• Online primal-dual inference algorithm for small-batch / streaming deployment with sample-weighted average violation guarantee (converges to ≤ ε as T → ∞)
• fairness_training.viz module: plot_training_history, plot_group_distributions, plot_fairness_tradeoff
• Quickstart Colab notebook (notebooks/quickstart.ipynb)
• Full MkDocs documentation site
• CI workflow (pytest on Python 3.10/3.11) and PyPI trusted-publishing workflow

Notes

• Protected attributes must be binary (0/1) columns in the feature matrix X
• Supports up to 2 protected attributes (marginal fairness, enforced independently)
• Training always uses hard per-batch constraints regardless of b_tau; b_tau controls the inference regime only (default: 64)
• weighted_avg_fairness_gap (the theorem-bounded quantity) ≠ pooled_fairness_gap; they coincide only when group ratios are constant across batches