Test doxygen

.gitignore

@@ -1,5 +1,6 @@
 /higher/data
 /higher/samples
+/higher/doc
 /Gradient-Descent-The-Ultimate-Optimizer/data
 /FAR-HO/data
 /__pycache__

higher/smart_aug/datasets.py

@@ -6,12 +6,17 @@ import torch
 from torch.utils.data import SubsetRandomSampler
 import torchvision
 
+#Train/Validation batch size.
 BATCH_SIZE = 300
-TEST_SIZE = 300 
+#Test batch size.
+TEST_SIZE = BATCH_SIZE 
 #TEST_SIZE = 10000 #legerement +Rapide / + Consomation memoire !
 
+#Wether to download data.
 download_data=False
+#Number of worker to use.
 num_workers=2 #4
+#Pin GPU memory
 pin_memory=False #True :+ GPU memory / + Lent
 
 #ATTENTION : Dataug (Kornia) Expect image in the range of [0, 1]
@@ -37,8 +42,10 @@ transform = torchvision.transforms.Compose([
 #)
 
 ### Classic Dataset ###
+#Training data
 data_train = torchvision.datasets.CIFAR10("../data", train=True, download=download_data, transform=transform)
 #data_val = torchvision.datasets.CIFAR10("../data", train=True, download=download_data, transform=transform)
+#Testing data
 data_test = torchvision.datasets.CIFAR10("../data", train=False, download=download_data, transform=transform)
 
 train_subset_indices=range(int(len(data_train)/2))

higher/smart_aug/model.py

@@ -5,7 +5,13 @@ import torch.nn.functional as F
 
 ## Basic CNN ##
 class LeNet(nn.Module):
+    """Basic CNN.
+
+    """
     def __init__(self, num_inp, num_out):
+        """Init LeNet.
+
+        """
         super(LeNet, self).__init__()
         self.conv1 = nn.Conv2d(num_inp, 20, 5)
         self.pool = nn.MaxPool2d(2, 2)
@@ -15,6 +21,9 @@ class LeNet(nn.Module):
         self.fc2 = nn.Linear(500, num_out)
 
     def forward(self, x):
+        """Main method of LeNet
+
+        """
         x = self.pool(F.relu(self.conv1(x)))
         x = self.pool2(F.relu(self.conv2(x)))
         x = x.view(x.size(0), -1)
@@ -23,4 +32,7 @@ class LeNet(nn.Module):
         return x
 
     def __str__(self):
+        """ Get name of model
+
+        """
         return "LeNet"

higher/smart_aug/transformations.py

@@ -18,15 +18,18 @@ import torch
 import kornia
 import random
 
+#TF that don't have use for magnitude parameter.
+TF_no_mag={'Identity', 'FlipUD', 'FlipLR', 'Random', 'RandBlend'}
+#TF which implemetation doesn't allow gradient propagaition.
+TF_no_grad={'Solarize', 'Posterize', '=Solarize', '=Posterize'}
+#TF for which magnitude should be ignored (Magnitude fixed).
+TF_ignore_mag= TF_no_mag | TF_no_grad 
 
-TF_no_mag={'Identity', 'FlipUD', 'FlipLR', 'Random', 'RandBlend'} #TF that don't have use for magnitude parameter.
+# What is the max 'level' a transform could be predicted
-TF_no_grad={'Solarize', 'Posterize', '=Solarize', '=Posterize'} #TF which implemetation doesn't allow gradient propagaition.
+PARAMETER_MAX = 1
-TF_ignore_mag= TF_no_mag | TF_no_grad #TF for which magnitude should be ignored (Magnitude fixed).
+# What is the min 'level' a transform could be predicted
+PARAMETER_MIN = 0.1 
 
-PARAMETER_MAX = 1  # What is the max 'level' a transform could be predicted
-PARAMETER_MIN = 0.1 # What is the min 'level' a transform could be predicted
-
-### Available TF for Dataug ###
 # Dictionnary mapping tranformations identifiers to their function.
 # Each value of the dict should be a lambda function taking a (batch of data, magnitude of transformations) tuple as input and returns a batch of data.
 TF_dict={ #Dataugv5+
@@ -416,13 +419,16 @@ def blend(x,y,alpha):
     return res
 
 #Not working
-def auto_contrast(x): #PAS OPTIMISE POUR DES BATCH #EXTRA LENT
+def auto_contrast(x):
-  # Optimisation : Application de LUT efficace / Calcul d'histogramme par batch/channel
+    """NOT TESTED - EXTRA SLOW
-  print("Warning : Pas encore check !")
+
-  (batch_size, channels, h, w) = x.shape
+    """
-  x = int_image(x) #Expect image in the range of [0, 1]
+    # Optimisation : Application de LUT efficace / Calcul d'histogramme par batch/channel
-  #print('Start',x[0])
+    print("Warning : Pas encore check !")
-  for im_idx, img in enumerate(x.chunk(batch_size, dim=0)): #Operation par image
+    (batch_size, channels, h, w) = x.shape
+    x = int_image(x) #Expect image in the range of [0, 1]
+    #print('Start',x[0])
+    for im_idx, img in enumerate(x.chunk(batch_size, dim=0)): #Operation par image
     #print(img.shape)
     for chan_idx, chan in enumerate(img.chunk(channels, dim=1)): # Operation par channel
       #print(chan.shape)
@@ -449,19 +455,22 @@ def auto_contrast(x): #PAS OPTIMISE POUR DES BATCH #EXTRA LENT
           chan[chan==ix]=n_ix
           x[im_idx, chan_idx]=chan
 
-  #print('End',x[0])
+    #print('End',x[0])
-  return float_image(x)
+    return float_image(x)
 
-def equalize(x): #PAS OPTIMISE POUR DES BATCH
+def equalize(x):
-  raise Exception(self, "not implemented") 
+    """ NOT WORKING
-  # Optimisation : Application de LUT efficace / Calcul d'histogramme par batch/channel
+
-  (batch_size, channels, h, w) = x.shape
+    """
-  x = int_image(x) #Expect image in the range of [0, 1]
+    raise Exception(self, "not implemented") 
-  #print('Start',x[0])
+    # Optimisation : Application de LUT efficace / Calcul d'histogramme par batch/channel
-  for im_idx, img in enumerate(x.chunk(batch_size, dim=0)): #Operation par image
+    (batch_size, channels, h, w) = x.shape
+    x = int_image(x) #Expect image in the range of [0, 1]
+    #print('Start',x[0])
+    for im_idx, img in enumerate(x.chunk(batch_size, dim=0)): #Operation par image
     #print(img.shape)
     for chan_idx, chan in enumerate(img.chunk(channels, dim=1)): # Operation par channel
       #print(chan.shape)
       hist = torch.histc(chan, bins=256, min=0, max=255) #PAS DIFFERENTIABLE
 
-  return float_image(x)
+    return float_image(x)