DataLoader processed in multiple remote computation machines for heavy data processing.
total_trainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=True, transform=transform_train)
loader = RemoteDataLoader(total_trainset, batch_size=32, timeout=5)
example.py
contains a simple example to process cifar10 images using remote nodes.
$ python example.py # run server(dataloader)
$ python remote_dataloader/worker.py --server {master_ip}:1958 # run multiple workers
$ python remote_dataloader/worker.py --server {master_ip}:1958
$ ...