Data loading and saving in different sparse and dense format.

multi_categorical_gans/datasets/formats.py

@@ -0,0 +1,38 @@
+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()