krypticmouse / MouseBit

An easy to use PyTorch utility for easy DataLoader creation.

MouseBit

An easy to use PyTorch utility for easy DataLoader creation. DataLoader creation can be redundant at times, keeping that in mind MouseBit was created to quickly create DataLoaders quickly.

New Features

• Added Text and Image Dataloader support.
• Wrapped Fetching Dataloader to a function instead of class.

Tutorial

Currently it only supports structured data but I'll be extending it's support for image and text data soon.

Creating Dataloaders for Structured Data

In order to get the dataloader start by creating the FetchDataLoader instance and calling get_dataloader method to get dataloader for the corresponding data.

>>> from mousebit.structured import FetchDataLoader
>>> msb = FetchDataLoader(X,Y)
>>> dataloader = msb.get_dataloader(self, batch_size = 64, drop_last = True, shuffle = True)

Getting Train and Test Dataloaders for Structured Data

To get train and test dataloaders MouseBit has get_dataloader_splits method in FetchDataLoader.

>>> from mousebit.structured import FetchDataLoader
>>> msb = FetchDataLoader(X,Y)
>>> trainloader, testloader = msb.get_dataloader_splits(self, batch_size = 64, stratify_on_y = False, random_state = 101, test_size = 0.3)