new functionality to change the similarity function

config.yaml

@@ -3,6 +3,6 @@ out_dim: 64
 s: 1
 temperature: 0.5
 base_convnet: "resnet18" # one of: "resnet18 or resnet50"
-use_cosine_similarity: True
+use_cosine_similarity: False
 epochs: 40
-num_workers: 4
+num_workers: 0

train.py

@@ -10,7 +10,7 @@ from torch.utils.tensorboard import SummaryWriter
 import torch.nn.functional as F
 
 from models.resnet_simclr import ResNetSimCLR
-from utils import get_negative_mask, get_augmentation_transform
+from utils import get_negative_mask, get_augmentation_transform, get_similarity_function
 
 torch.manual_seed(0)
 
@@ -23,14 +23,17 @@ use_cosine_similarity = config['use_cosine_similarity']
 
 data_augment = get_augmentation_transform(s=config['s'], crop_size=96)
 
-train_dataset = datasets.STL10('./data', split='train', download=True, transform=transforms.ToTensor())
+train_dataset = datasets.STL10('./data', split='train+unlabeled', download=True, transform=transforms.ToTensor())
+# train_dataset = datasets.Caltech101(root='./data', target_type="category", transform=transforms.ToTensor(),
+#                                     target_transform=None, download=True)
+
 train_loader = DataLoader(train_dataset, batch_size=batch_size, num_workers=config['num_workers'], drop_last=True,
                           shuffle=True)
 
 # model = Encoder(out_dim=out_dim)
 model = ResNetSimCLR(base_model=config["base_convnet"], out_dim=out_dim)
 
-train_gpu = False  # torch.cuda.is_available()
+train_gpu = torch.cuda.is_available()
 print("Is gpu available:", train_gpu)
 
 # moves the model parameters to gpu
@@ -42,9 +45,7 @@ optimizer = optim.Adam(model.parameters(), 3e-4)
 
 train_writer = SummaryWriter()
 
-if use_cosine_similarity:
-    cos_similarity_dim1 = torch.nn.CosineSimilarity(dim=1)
-    cos_similarity_dim2 = torch.nn.CosineSimilarity(dim=2)
+sim_func_dim1, sim_func_dim2 = get_similarity_function(use_cosine_similarity)
 
 # Mask to remove positive examples from the batch of negative samples
 negative_mask = get_negative_mask(batch_size)
@@ -86,13 +87,7 @@ for e in range(config['epochs']):
         # assert zis.shape == (batch_size, out_dim), "Shape not expected: " + str(zis.shape)
         # assert zjs.shape == (batch_size, out_dim), "Shape not expected: " + str(zjs.shape)
 
-        # positive pairs
-        if use_cosine_similarity:
-            l_pos = cos_similarity_dim1(zis.view(batch_size, out_dim), zjs.view(batch_size, out_dim)).view(batch_size,
-                                                                                                           1)
-        else:
-            l_pos = torch.bmm(zis.view(batch_size, 1, out_dim), zjs.view(batch_size, out_dim, 1)).view(batch_size, 1)
-
+        l_pos = sim_func_dim1(zis, zjs).view(batch_size, 1)
         l_pos /= temperature
         # assert l_pos.shape == (batch_size, 1), "l_pos shape not valid" + str(l_pos.shape)  # [N,1]
 
@@ -101,13 +96,7 @@ for e in range(config['epochs']):
         loss = 0
 
         for positives in [zis, zjs]:
-
-            if use_cosine_similarity:
-                l_neg = cos_similarity_dim2(positives.view(batch_size, 1, out_dim),
-                                            negatives.view(1, (2 * batch_size), out_dim))
-            else:
-                l_neg = torch.tensordot(positives.view(batch_size, 1, out_dim),
-                                        negatives.T.view(1, out_dim, (2 * batch_size)), dims=2)
+            l_neg = sim_func_dim2(positives, negatives)
 
             labels = torch.zeros(batch_size, dtype=torch.long)
             if train_gpu:

utils.py

@@ -21,9 +21,10 @@ def get_negative_mask(batch_size):
 
 class GaussianBlur(object):
     # Implements Gaussian blur as described in the SimCLR paper
-    def __init__(self, min=0.1, max=2.0, kernel_size=9):
+    def __init__(self, kernel_size, min=0.1, max=2.0):
         self.min = min
         self.max = max
+        # kernel size is set to be 10% of the image height/width
         self.kernel_size = kernel_size
 
     def __call__(self, sample):
@@ -43,22 +44,28 @@ def get_augmentation_transform(s, crop_size):
     # get a set of data augmentation transformations as described in the SimCLR paper.
     color_jitter = transforms.ColorJitter(0.8 * s, 0.8 * s, 0.8 * s, 0.2 * s)
     data_aug_ope = transforms.Compose([transforms.ToPILImage(),
-                                       transforms.RandomResizedCrop(crop_size),
+                                       transforms.RandomResizedCrop(size=crop_size),
                                        transforms.RandomHorizontalFlip(),
                                        transforms.RandomApply([color_jitter], p=0.8),
                                        transforms.RandomGrayscale(p=0.2),
-                                       GaussianBlur(),
+                                       GaussianBlur(kernel_size=int(0.1 * crop_size)),
                                        transforms.ToTensor()])
     return data_aug_ope
 
 
 def _dot_simililarity_dim1(x, y):
-    v = torch.bmm(x.unsqueeze(1), y.unsqueeze(2))
+    # x shape: (N, 1, C)
+    # y shape: (N, C, 1)
+    # v shape: (N, 1, 1)
+    v = torch.bmm(x.unsqueeze(1), y.unsqueeze(2))  #
     return v
 
 
 def _dot_simililarity_dim2(x, y):
     v = torch.tensordot(x.unsqueeze(1), y.T.unsqueeze(0), dims=2)
+    # x shape: (N, 1, C)
+    # y shape: (1, C, 2N)
+    # v shape: (N, 2N)
     return v
 
 
@@ -68,6 +75,9 @@ def _cosine_simililarity_dim1(x, y):
 
 
 def _cosine_simililarity_dim2(x, y):
+    # x shape: (N, 1, C)
+    # y shape: (1, 2N, C)
+    # v shape: (N, 2N)
     v = cos2d(x.unsqueeze(1), y.unsqueeze(0))
     return v