Trained model on a notebook

[SaiChamarty]

Added training loop and trained over 50 epochs for images with original resolution - 512x512

This is implemented on u-net-initial-implementation.ipynb

Modified the following:

changed transforms to unet_transforms so that it does not conflict with the original transforms function from torchvision.
changed crop size to 512x512 instead of 256x256 and removed additional redundant crops, as we did not need it for 512.
changed number of channels to 1 instead of 3 in the unet definition. Our dataset does not have a color scheme - they are grayscale.
added weight initialization like on the paper - He initialization
used dice score and cross entropy loss instead of only cross entropy. Dice score is used on the paper.
To run, you would have to do the following:

have an env with all the existing dependencies shown in the first cell of the notebook.
I used pip as I felt it was more native for me, but I believe you could use uv as well.