Quickstart

Install

pip install -e .

Train a predictor

python -m src.tabstruct.experiment.run_experiment \
  --model mlp \
  --dataset adult \
  --test_size 0.2 \
  --valid_size 0.1 \
  --tags dev

Evaluate a saved checkpoint

python -m src.tabstruct.experiment.run_experiment \
  --model mlp \
  --eval_only \
  --saved_checkpoint_path /path/to/checkpoint \
  --dataset adult \
  --test_size 0.2 \
  --valid_size 0.1 \
  --tags dev

Generate and evaluate synthetic data

# Generate
python -m src.tabstruct.experiment.run_experiment \
  --pipeline generation \
  --generation_only \
  --generation_ratio 10 \
  --model smote \
  --dataset mfeat-fourier \
  --test_size 0.2 \
  --valid_size 0.1 \
  --tags dev

# Evaluate past generation
python -m src.tabstruct.experiment.run_experiment \
  --pipeline generation \
  --model smote \
  --eval_only \
  --dataset mfeat-fourier \
  --test_size 0.2 \
  --valid_size 0.1 \
  --generator_tags dev \
  --curate_ratio 10 \
  --tags dev