Skip to main content

Module data

Module data 

Source

Re-exports§

pub use dataset::Batch;
pub use dataset::Dataset;
pub use loader::DataLoader;
pub use mmap::MmapDataset;
pub use sharded::ShardedDataset;

Modules§

collate
Batch collation and index shuffling helpers.
dataset
Dataset trait and batch type for training data pipelines.
loader
DataLoader with shuffling, batching, and background prefetching.
mmap
Memory-mapped dataset for large token files.
sharded
Sharded dataset wrapper for distributed training.