build_synthetic_cifar10_dataloader#

composer.datasets.build_synthetic_cifar10_dataloader(global_batch_size, is_train=True, drop_last=True, shuffle=True, num_unique_samples=100, device='cpu', memory_format=MemoryFormat.CONTIGUOUS_FORMAT, **dataloader_kwargs)[source]#

Builds a synthetic CIFAR-10 dataset for debugging or profiling.

Parameters

  • global_batch_size (int) โ€“ Global batch size

  • is_train (bool) โ€“ Whether to load the training data or validation data. Default: True.

  • drop_last (bool) โ€“ Drop remainder samples. Default: True.

  • shuffle (bool) โ€“ Shuffle the dataset. Default: True.

  • num_unique_samples (int) โ€“ number of unique samples in synthetic dataset. Default: 100.

  • device (str) โ€“ device with which to load the dataset. Default: cpu.

  • memory_format (composer.core.MemoryFormat) โ€“ memory format of the tensors. Default: CONTIGUOUS_FORMAT.

  • **dataloader_kwargs (Any) โ€“ Additional settings for the dataloader (e.g. num_workers, etc.)