How to split a dataset into multiple databases and create a new dataset

Hi all,

I have an image dataset loaded from ImageFolder:

import torchvision.datasets as datasets

train_dataset = datasets.ImageFolder(root=train_dir, transform=transform)

I want to split image into multiple chunk and create a new train_dataset, something like: 
new_train_dataset = torch.utils.data.TensorDataset(img_chunk1, img_chunk2, img_chunk3, y_train)

then create train_loader from the new dataset with multiple inputs to train model as: 

for i, (input1, input2, input3, labels) in enumerate(train_load):
     .....
     output = model(input1, input2, input3) 

How to split the first train_dataset to small chunk of images and feed into new_train_dataset?

Thank you very much for help. @ptrblck