python source code of formats

Project: multi-categorical-gans (GitHub Link)

multi-categorical-gans-master
- LICENSE
- README.md
- requirements.txt
- .gitignore
- multi_categorical_gans
  - datasets
    - train_test_split.py
    - synthetic
      - generate.py
      - __init__.py
      - README.md
    - fold_split.py
    - __init__.py
    - README.md
    - formats.py
    - dataset.py
    - uscensus
      - transform.py
      - reverse_transform.py
      - __init__.py
      - README.md
  - utils
    - cuda.py
    - commandline.py
    - prediction.py
    - logger.py
    - initialization.py
    - __init__.py
    - categorical.py
  - __init__.py
  - methods
    - arae
      - sampler.py
      - __init__.py
      - README.md
      - trainer.py
    - medgan
      - sampler.py
      - pre_trainer.py
      - generator.py
      - discriminator.py
      - __init__.py
      - README.md
      - trainer.py
    - mc_wgan_gp
      - sampler.py
      - __init__.py
      - README.md
      - trainer.py
    - __init__.py
    - README.md
    - general
      - decoder.py
      - encoder.py
      - generator.py
      - discriminator.py
      - __init__.py
      - multi_categorical.py
      - autoencoder.py
      - single_output.py
      - wgan_gp.py
    - mc_gumbel
      - sampler.py
      - __init__.py
      - README.md
      - trainer.py
  - metrics
    - mse_predictions_by_dimension.py
    - mse_predictions_by_categorical.py
    - __init__.py
    - README.md
    - mse_probabilities_by_dimension.py

import numpy as np

from scipy.sparse import load_npz, save_npz


def load_dense(features_path, transform=True):
    features = np.load(features_path)
    if transform:
        features = features.astype(np.float32)
    return features


def load_sparse(features_path, transform=True):
    features = load_npz(features_path)
    if transform:
        features = np.asarray(features.todense()).astype(np.float32)
    return features


def save_dense(features_path, features):
    np.save(features_path, features)


def save_sparse(features_path, features):
    save_npz(features_path, features)


loaders = {
    "dense": load_dense,
    "sparse": load_sparse,
}

savers = {
    "dense": save_dense,
    "sparse": save_sparse,
}

data_formats = loaders.keys()