ML Katas

Create a custom torch.utils.data.Dataset that doesn't load data from disk. Instead, the __getitem__ method should generate an image on the fly (e.g., a simple geometric shape, a random pattern, or even a fractal). This is a useful skill for when you don't have a pre-existing dataset. The __len__ method should simply return a large number of samples, and __getitem__ should create a new tensor with the image and a corresponding label for each call.

Verification: Iterate through your DataLoader and visualize a few of the generated images. Each image in the batch should be unique and should correspond to the logic you defined in __getitem__, proving that the generation is working correctly.