udpate

Author: qnguyen3

SimCLR.py

@@ -137,7 +137,7 @@ def init_model(self):
 
         elif self.arch == 'resnet50':
             # backbone = resnet.resnet50(mode=self.mode)
-            backbone = resnet.ResNetPreTrained(type='resnet50')
+            backbone = resnet.resnet50(type='resnet50')
         return backbone
 
     def forward(self, x):

module/resnet.py

@@ -78,7 +78,7 @@ def __init__(self, block, num_blocks, in_channel=3, zero_init_residual=False, mo
         super(ResNet, self).__init__()
         self.in_planes = 64
         self.mode = mode
-        if self.mode == 'cifar':
+        if self.mode == 'cifar10':
             self.conv1 = nn.Conv2d(in_channel, 64, kernel_size=3, stride=1, padding=1,
                                 bias=False)
             self.bn1 = nn.BatchNorm2d(64)
@@ -125,8 +125,8 @@ def _make_layer(self, block, planes, num_blocks, stride):
 
     def forward(self, x):
         out = F.relu(self.bn1(self.conv1(x)))
-        if self.mode == 'cifar':
-            out = self.maxpool(out)
+        # if self.mode == 'cifar10':
+        #     out = self.maxpool(out)
         out = self.layer1(out)
         out = self.layer2(out)
         out = self.layer3(out)

train_cifar10.py

@@ -17,7 +17,7 @@
 train_data = CIFAR10(download=True,root="./cifar10",transform=MultiViewDataInjector([train_transform,train_transform,val_test_transform]))
 train_len = len(train_data)
 num_class = len(np.unique(train_data.targets))
-train_loader = DataLoader(dataset = train_data, batch_size = 1024, num_workers=48)
+train_loader = DataLoader(dataset = train_data, batch_size = 1024, num_workers=12)
 # test_loader = DataLoader(dataset = test_data, batch_size = 16)
 # valid_loader = DataLoader(dataset = val_data, batch_size= 16)