Add support for loading .npy dataset in training pipeline

[shrutiishar]

Title: Proposal: Add support for .npy dataset loading in training pipeline

Hi,

I’ve been exploring the DeepLense dataset pipeline and noticed that the data is stored in .npy format, which is not directly compatible with torchvision’s ImageFolder.

I implemented a custom PyTorch Dataset class to load .npy files and convert them into tensors suitable for CNN models like ResNet.

This could make it easier to train models directly on simulated lensing data without additional preprocessing.

Would this be useful to integrate into the project? I’d be happy to contribute this.

Thanks!

[shrutiishar]

Hi, I’ve created a PR for this. Would appreciate your feedback.