Home Knowledge Base multi-GPU ML

multi-GPU ML

No mentions found

This entity hasn't been tracked yet, or Iris is still building its knowledge base.

Related Articles from SNS

Resource-aware Computation-Communication Overlap for multi-GPU ML Workloads

new Abstract: The rapid growth of large-scale machine learning (ML) has made distributed training across multiple GPUs a fundamental component of modern ML systems. As model sizes and computational throughput continue to increase, communication overhead has become a dominant bottleneck in multi-GPU training, particularly when computation and communication are executed sequentially. This work explores concurrent execution of computation and collective communication using two portable runtime...

arXiv CS 1d ago