Development

cortex/datasets/__init__.py

@@ -42,12 +42,16 @@ def resolve(c):
     from .basic.caltech import CALTECH
     from .basic.uci import UCI
     from .basic.cifar import CIFAR
+    from .basic.europarl import Europarl
+    from .basic.voc import VOC
 
     r_dict = {
         'mnist': MNIST,
         'cifar': CIFAR,
         'caltech': CALTECH,
-        'uci': UCI
+        'uci': UCI,
+        'europarl': Europarl,
+        'voc': VOC
     }
 
     C = r_dict.get(c, None)
@@ -70,7 +74,7 @@ def make_one_hot(Y, n_classes=None):
         class_list = np.unique(Y).tolist()
         n_classes = len(class_list)
     else:
-        class_list = range(n_classes)
+        class_list = range(0, n_classes)
 
     if Y.ndim == 2:
         reshape = Y.shape
@@ -86,7 +90,7 @@ def make_one_hot(Y, n_classes=None):
             i = class_list.index(Y[idx])
         except ValueError:
             raise ValueError('Class list is missing elements')
-        O[idx, i] = 1.;
+        O[idx, i] = 1.
 
     if reshape is not None:
         O = O.reshape(reshape + (n_classes,))
@@ -176,7 +180,7 @@ def dataset_factory(resolve_dataset, dataset=None, split=[0.7, 0.2, 0.1],
             valid_batch_size=valid_batch_size, test_batch_size=test_batch_size,
             **dataset_args)
     else:
-        train, valid, test, idx =  C.factory(
+        train, valid, test, idx = C.factory(
             split=split, idx=idx,
             batch_sizes=[train_batch_size, valid_batch_size, test_batch_size],
             **dataset_args)
@@ -367,6 +371,7 @@ def __init__(self, data, distributions=None, labels='label', name=None,
         self.balance = balance
 
         self.dims = dict()
+        self.dimsall = dict()
         if distributions is None:
             self.distributions = dict()
         else:
@@ -391,6 +396,7 @@ def __init__(self, data, distributions=None, labels='label', name=None,
                                     'number of samples (shape[0]), '
                                     '(%d vs %d)' % (self.n, v.shape[0]))
             self.dims[k] = v.shape[1]
+            self.dimsall[k] = v.shape
             if not k in self.distributions.keys():
                 self.distributions[k] = 'binomial'