Can I read Save: Sparsity-Aware Vector Engine For Accelerating DNN Training and Inference On Cpus on EtoBox?

Save: Sparsity-Aware Vector Engine For Accelerating DNN Training and Inference On Cpus by giddammagiddi is a document available to read on EtoBox.

What is Save: Sparsity-Aware Vector Engine For Accelerating DNN Training and Inference On Cpus about?

The document presents SAVE, a novel vector engine designed for CPUs to enhance the efficiency of Deep Neural Network (DNN) training and inference by exploiting unstructured sparsity in General Matrix Multiplication (GEMM). SAVE accelerates performance by dynamically skipping ineffectual computations and optimizing resource usage, achieving speed-ups of 1.37x-1.68x in inference and 1.28x-1.64x in training across various DNN models. The proposed architecture is transparent to software, making it applicable to

Author

giddammagiddi

Language

EN