style: reduce bit misspell code comments README

README.md

@@ -108,7 +108,7 @@ For unstructured pruning, we strictly follow the ones adopted by the state of th
 k = 0.8000 0.6400 0.5120 0.4100 0.3280 0.2620 0.2097 0.1678 0.1342 0.1074 0.0859 0.0687 0.0550 0.0440 0.0350 0.0280 0.0225 0.0180 0.0144 0.0115 0.0092
 ```
 
-Please do not be misled by this sparsity series. For each sparse ratio alone, the BiP prunes from dense model directly to the target sparsity ratio, rather than pruning based on the last checkpoint iteratively. 
+Please do not be mislead by this sparsity series. For each sparse ratio alone, the BiP prunes from dense model directly to the target sparsity ratio, rather than pruning based on the last checkpoint iteratively. 
 
 For structured pruning, we follow the linear sparsity ratio series (Figure 4, A6, and A7):
 

args.py

@@ -51,7 +51,7 @@ def parse_args():
         "--scaled-score-init",
         action="store_true",
         default=False,
-        help="Init importance scores proportaional to weights (default kaiming init)",
+        help="Init importance scores proportional to weights (default kaiming init)",
     )
 
     parser.add_argument(

models/basic.py

@@ -8,7 +8,7 @@ def forward(self, input):
         return input.view(input.size(0), -1)
 
 
-# lin_i: i layer linear feedforard network.
+# lin_i: i layer linear feedforward network.
 def lin_1(input_dim=3072, num_classes=10):
     model = nn.Sequential(nn.Flatten(), nn.Linear(input_dim, num_classes))
     return model

models/resnet.py

@@ -214,7 +214,7 @@ def forward(self, x):
         return self._forward_impl(x)
 
 
-# NOTE: Only supporting default (kaiming_init) initializaition.
+# NOTE: Only supporting default (kaiming_init) initialization.
 def ResNet18(conv_layer, linear_layer, **kwargs):
     return ResNet(conv_layer, linear_layer, BasicBlock, [2, 2, 2, 2], **kwargs)
 

models/resnets.py

@@ -4,7 +4,7 @@
 which is implemented for ImageNet and doesn't have option A for identity.
 Moreover, most of the implementations on the web is copy-paste from
 torchvision's resnet and has wrong number of params.
-Proper ResNet-s for CIFAR10 (for fair comparision and etc.) has following
+Proper ResNet-s for CIFAR10 (for fair comparison and etc.) has following
 number of layers and parameters:
 name      | layers | params
 ResNet20  |    20  | 0.27M

models/wrn_cifar.py

@@ -162,7 +162,7 @@ def forward(self, x):
         return self.fc(out)
 
 
-# NOTE: Only supporting default (kaiming_init) initializaition.
+# NOTE: Only supporting default (kaiming_init) initialization.
 def wrn_28_10(conv_layer, linear_layer, **kwargs):
     return WideResNet(conv_layer, linear_layer, depth=28, widen_factor=10, **kwargs)
 

train.py

@@ -39,7 +39,7 @@ def main():
     if args.exp_mode in ["prune", "finetune"] and not args.resume:
         assert args.source_net, "Provide checkpoint to prune/finetune"
 
-    # create resutls dir (for logs, checkpoints, etc.)
+    # create results dir (for logs, checkpoints, etc.)
     result_main_dir = os.path.join(Path(args.result_dir), args.exp_name, args.exp_mode)
 
     if os.path.exists(result_main_dir):