Update autograd_tutorial.py (#2021)

The parameter 'pretrained' is deprecated. Updated it to latest standard.

Co-authored-by: Nikita Shulga <[email protected]>

Author: AdityaPunetha

beginner_source/blitz/autograd_tutorial.py

@@ -43,8 +43,9 @@
     This tutorial work only on CPU and will not work on GPU (even if tensor are moved to CUDA).
 
 """
-import torch, torchvision
-model = torchvision.models.resnet18(pretrained=True)
+import torch
+from torchvision.models import resnet18, ResNet18_Weights
+model = resnet18(weights=ResNet18_Weights.DEFAULT)
 data = torch.rand(1, 3, 64, 64)
 labels = torch.rand(1, 1000)
 
@@ -283,7 +284,7 @@
 
 from torch import nn, optim
 
-model = torchvision.models.resnet18(pretrained=True)
+model = resnet18(weights=ResNet18_Weights.DEFAULT)
 
 # Freeze all the parameters in the network
 for param in model.parameters():