For the lesson7, the restoring image part, did it use a pre-trained model?

Part 1 (2019)
1

It looks like it use pre-trained vgg16 image net. If I have my own pth that was a resnet 34 pre-trained, how do I load the pre-train?