Merge pull request #108 from joeljpoulin/tox-additions

Tox additions & Autopep8 ran once.

cortex/_lib/data/__init__.py

@@ -19,7 +19,6 @@
 def setup(source: str=None, batch_size=64, n_workers: int=4,
           skip_last_batch: bool=False, DataLoader=None,
           copy_to_local: bool=False, data_args={}, shuffle: bool=True):
-
     """
     Dataset entrypoint.
 

cortex/_lib/train.py

@@ -127,7 +127,7 @@ def display_results(
     if times:
         print('\tAvg update times: ' + ' | '
               .join(['{}: {:.2f}'
-                    .format(k, v) for k, v in times.items()]))
+                     .format(k, v) for k, v in times.items()]))
 
     train_losses = train_results.pop('losses')
     test_losses = test_results.pop('losses')
@@ -142,15 +142,15 @@ def display_results(
             if isinstance(v_train, dict):
                 print('\t' + k + ': ' + ' | '
                       .join(['{}: {:.2f}'
-                            .format(k_, v_train[k_])
+                             .format(k_, v_train[k_])
                              for k_ in v_train.keys()]))
             else:
                 print('\t{}: {:.2f}'.format(k, v_train))
         else:
             if isinstance(v_train, dict):
                 print('\t' + k + ': ' + ' | '
                       .join(['{}: {:.2f} / {:.2f}'
-                            .format(k_, v_train[k_], v_test[k_])
+                             .format(k_, v_train[k_], v_test[k_])
                              for k_ in v_train.keys()]))
             else:
                 print('\t{}: {:.2f} / {:.2f}'.format(k, v_train, v_test))

cortex/built_ins/models/classifier.py

@@ -58,6 +58,8 @@ def classify(self, classifier, inputs, targets, criterion=None):
     def visualize(self, inputs, targets, predicted):
         self.add_image(inputs.data, labels=(targets.data, predicted.data),
                        name=self.name + '_gt_pred')
+
+
 register_plugin(ClassifyRoutine)
 
 
@@ -83,6 +85,8 @@ def build(self, dim_in: int=None, dim_h=[200, 200], classifier_args={}):
         classifier = FullyConnectedNet(dim_in, dim_h=dim_h,
                                        dim_out=dim_l, **classifier_args)
         self.add_networks(simple_classifier=classifier)
+
+
 register_plugin(SimpleClassifierBuild)
 
 
@@ -113,6 +117,8 @@ def build(self, classifier_type='convnet', classifier_args={}):
 
         classifier = Encoder(shape, dim_out=dim_l, **args)
         self.add_networks(image_classifier=classifier)
+
+
 register_plugin(ImageClassifierBuild)
 
 
@@ -135,4 +141,6 @@ def __init__(self):
                          inputs='data.images', targets='data.targets',
                          images='data.images')
         self.add_train_procedure('classification')
+
+
 register_plugin(ImageClassification)

cortex/built_ins/models/gan.py

@@ -145,6 +145,8 @@ def run(self, measure: str='GAN'):
                                 real=P_samples.view(-1).data),
                            name='discriminator output')
         self.losses.discriminator = -difference
+
+
 register_plugin(DiscriminatorRoutine)
 
 
@@ -209,7 +211,9 @@ def get_gradient(inp, output):
 
             try:
                 epsilon = self.inputs.e.view(-1, 1, 1, 1)
+
             except AttributeError:
+
                 raise ValueError('You must initiate a uniform random variable'
                                  '`e` to use interpolation')
             mid_in = ((1. - epsilon) * inp1 + epsilon * inp2)
@@ -226,6 +230,8 @@ def get_gradient(inp, output):
                 'Unsupported penalty {}'.format(penalty_type))
 
         return penalty_amount * penalty
+
+
 register_plugin(PenaltyRoutine)
 
 
@@ -263,6 +269,8 @@ def run(self, measure: str=None, loss_type: str='non-saturating'):
         self.add_image(X_Q, name='generated')
 
         return X_Q
+
+
 register_plugin(GeneratorRoutine)
 
 
@@ -288,6 +296,8 @@ def build(self, discriminator_type: str='convnet', discriminator_args={}):
             encoder_args=discriminator_args)
         discriminator = Encoder(x_shape, dim_out=1, **discriminator_args)
         self.add_networks(discriminator=discriminator)
+
+
 register_plugin(DiscriminatorBuild)
 
 
@@ -298,6 +308,7 @@ class GeneratorBuild(BuildPlugin):
     plugin_name = 'generator'
     plugin_nets = ['generator']
 
+
     def build(self, generator_noise_type='normal', dim_z=64, generator_type: str='convnet',
               generator_args={}):
         '''
@@ -318,6 +329,8 @@ def build(self, generator_noise_type='normal', dim_z=64, generator_type: str='co
         generator = Decoder(x_shape, dim_in=dim_z, **generator_args)
 
         self.add_networks(generator=generator)
+
+
 register_plugin(GeneratorBuild)
 
 
@@ -345,4 +358,6 @@ def __init__(self):
         self.add_train_procedure(
             'generator', 'discriminator', 'gradient_penalty')
         self.add_eval_procedure('generator', 'discriminator')
+
+
 register_plugin(GAN)

cortex/built_ins/models/utils.py

@@ -52,7 +52,7 @@ def ms_ssim(X_a, X_b, window_size=11, size_average=True, C1=0.01**2, C2=0.03**2)
     def gaussian(sigma=1.5):
         gauss = torch.Tensor(
             [math.exp(-(x - window_size // 2) **
-             2 / float(2 * sigma ** 2)) for x in range(window_size)])
+                      2 / float(2 * sigma ** 2)) for x in range(window_size)])
         return gauss / gauss.sum()
 
     def create_window():

cortex/built_ins/models/vae.py

@@ -28,6 +28,7 @@ class VAENetwork(nn.Module):
         latent: The latent state (Z).
 
     '''
+
     def __init__(self, encoder, decoder, dim_out=None, dim_z=None):
         super(VAENetwork, self).__init__()
         self.encoder = encoder
@@ -96,6 +97,8 @@ def run(self, vae_criterion=F.mse_loss, beta_kld=1.):
         self.add_scatter(vae_net.mu.data, labels=Y.data, name='latent values')
 
         return vae_net.mu
+
+
 register_plugin(VAERoutine)
 
 
@@ -121,6 +124,8 @@ def build(self, encoder_type: str='convnet', dim_out: int=64, encoder_args={}):
                                                     encoder_args=encoder_args)
         encoder = Encoder(x_shape, dim_out=dim_out, **encoder_args)
         self.add_networks(image_encoder=encoder)
+
+
 register_plugin(ImageEncoderBuild)
 
 
@@ -146,6 +151,8 @@ def build(self, decoder_type: str='convnet', dim_in: int=64, decoder_args={}):
                                                     decoder_args=decoder_args)
         decoder = Decoder(x_shape, dim_in=dim_in, **decoder_args)
         self.add_networks(image_decoder=decoder)
+
+
 register_plugin(ImageDecoderBuild)
 
 
@@ -173,6 +180,8 @@ def build(self, dim_z=64, dim_encoder_out=1028):
         vae = VAENetwork(encoder, decoder, dim_out=dim_encoder_out,
                          dim_z=dim_z)
         self.add_networks(vae=vae)
+
+
 register_plugin(VAEBuild)
 
 
@@ -215,4 +224,6 @@ def __init__(self, add_classification=True):
             self.add_train_procedure('VAE', 'classification')
         else:
             self.add_train_procedure('VAE')
+
+
 register_plugin(VAE)

cortex/built_ins/networks/fully_connected.py

@@ -13,8 +13,10 @@
 
 
 class FullyConnectedNet(nn.Module):
+
     def __init__(self, dim_in, dim_out=None, dim_h=64, nonlinearity='ReLU', n_levels=None,
                  **layer_args):
+
         super(FullyConnectedNet, self).__init__()
         models = nn.Sequential()
 

cortex/built_ins/networks/resnets.py

@@ -28,7 +28,7 @@ def __init__(self, dim_in, dim_out, f_size, nonlinearity=None,
         models.add_module(name, Conv2d(dim_in, dim_out, f_size, 1, 1,
                                        bias=False))
         models.add_module(name + '_pool', nn.AvgPool2d(2,
-                          count_include_pad=False))
+                                                       count_include_pad=False))
         if nonlinearity:
             models.add_module('{}_{}'.format(
                 name, nonlinearity.__class__.__name__), nonlinearity)
@@ -80,7 +80,7 @@ def __init__(self, dim_in, dim_out, f_size, nonlinearity=None, prefix='', spectr
 
         models.add_module(name + '_up', nn.Upsample(scale_factor=2))
         models.add_module(name, Conv2d(dim_in, dim_out, f_size, 1, 1,
-                          bias=False))
+                                       bias=False))
 
         if nonlinearity:
             models.add_module(
@@ -243,7 +243,7 @@ def __init__(self, shape, dim_out=None, dim_h=64,
 
         name = 'conv_({}/{})_0'.format(dim_in, dim_out)
         models.add_module(name, Conv2d(dim_in, dim_out, f_size, 1, 1,
-                          bias=False))
+                                       bias=False))
 
         dim_out = dim_h
         for i in range(n_steps):

tox.ini

@@ -50,4 +50,18 @@ skip_install = false
 deps =
     pylint == 1.8.1
 commands =
-    pylint cortex cortex/_lib cortex/built_ins
+    pylint cortex cortex/_lib cortex/built_ins
+
+[testenv:autopep8]
+deps = 
+    autopep8
+commands =
+    autopep8 -a -i -r cortex
+
+[testenv:docs]
+basepython = python
+changedir = docs
+deps = sphinx
+commands =
+  sphinx-apidoc -f -o source ../cortex
+  make html