Index
optimus_dl.modules.data.datasets
¶
Modules and Sub-packages¶
base: Base dataset class for data sources.composite: Generate a unique seed for each rank based on the base seed, rank, and epoch.huggingface: Configuration for Hugging Face datasets.loop_dataset: Dataset that infinitely loops over an inner dataset.strategies:tokenized_dataset: Configuration for pre-tokenized sharded datasets.tokenized_flat_dataset: Configuration for flat tokenized datasets.txt_lines: Configuration for line-based text datasets.