Mode.train() question

When the model is in model.train(), and if we just keep passing image through model

for i, data in enumerate(data_loader):
x,y = data
f = model(x)

Is question the model learning high activation features?


I have a set of 1k images, and I want the model to look at the particular subject which is in every single image, there is no such thing as a label. The process is pure unsupervised.

How can I make the model learn the features of the subject seen in every image?
Thank you