gigl.experimental.knowledge_graph_embedding.lib.config.dataloader

Classes

DataloaderConfig

Configuration for PyTorch DataLoader used in knowledge graph embedding training.

Module Contents

class gigl.experimental.knowledge_graph_embedding.lib.config.dataloader.DataloaderConfig

Configuration for PyTorch DataLoader used in knowledge graph embedding training.

Controls data loading efficiency and memory usage during training and evaluation.

num_workers

Number of worker processes for data loading. Higher values can improve data loading speed but use more memory and CPU cores. Setting to 0 uses the main process for data loading. Defaults to 1.

Type:

int

pin_memory

Whether to pin loaded data tensors to GPU memory for faster host-to-device transfer. Should be True when using CUDA for training. May use additional GPU memory. Defaults to True.

Type:

bool

num_workers: int = 1

pin_memory: bool = True