Majid Al-RaimiPhD notes / KFUPM

COE 592/Lecture 4.2Pruning ratios, fine-tuning and sparse hardware

Overview
Full guide
Glossary
Reference
Practice quiz

Parts13

Where we are: the pruning problem so far3 concepts, slides 1-7
Why every layer needs its own pruning ratio3 concepts, slides 8-17
From sensitivity curves to per-layer ratios3 concepts, slides 18-23
AMC: pruning ratios as a reinforcement learning problem4 concepts, slides 24-32
NetAdapt: platform-aware pruning to a latency budget2 concepts, slides 33-40
Fine-tuning, iterative pruning and regularization4 concepts, slides 41-52
EIE: the first accelerator for sparse, compressed networks3 concepts, slides 53-60
Inside EIE: dataflow, the PE and its storage formats3 concepts, slides 61-70
EIE results and what came after5 concepts, slides 71-77
M:N sparsity on NVIDIA tensor cores4 concepts, slides 78-83
Sparse inputs and sparse convolution3 concepts, slides 84-96
TorchSparse: regular computation from irregular sparsity3 concepts, slides 97-109
PointAcc, the lecture summary and references4 concepts, slides 110-115

Majid Al-RaimiPractice quiz

COE 592Lecture 4.2Practice quiz

Practice quiz

There are no practice questions for this lecture yet.