Tag: distributed training

Researchers propose CROG and CLIG, two aggregation algorithms that weight client updates by Euclidean gradient distances to stabilize federated learning ...

A new survey from the National University of Defense Technology systematically compares distributed parallelism and memory-saving techniques that make large ...

Researchers have developed pFedDS, a personalized federated learning approach that measures each model parameter's sensitivity to non-uniform data and adapts ...

A new open-source simulator called FLInterrupt makes sudden client dropouts a controllable experimental variable in federated learning research, revealing how ...

Researchers at the National University of Defense Technology have systematically surveyed memory-efficient training techniques for large Transformer models, showing how ...

Researchers have developed Fed-A²LoRA, a federated fine-tuning method that halves trainable adapter parameters and cuts communication costs by up to ...

A comprehensive survey explains how accelerators, networks, storage, parallelism and fault tolerance are being redesigned to train increasingly large language ...