search

XtractOpen / Meganet.jl

A fresh approach to deep learning written in Julia
http://www.xtract.ai/
Other
14 stars 9 forks source link

batch norm optimized #40

Closed DavidBegert closed 6 years ago

DavidBegert commented 6 years ago

30% faster for apply, 75% faster for JYmv, 55% faster for JYTmv 38% mem improvement for JYmv, 40% Mem improvement for JYTmv

coveralls commented 6 years ago

Coverage Status

Coverage decreased (-0.02%) to 78.115% when pulling a76533d75cf74b1d1a2b7eae552fc1835e42aedb on DavidBegert:pull-request/c8027c9a into 547478bf4d02ffdf5b66aa5ba79fa36d86c9637f on XtractOpen:dev.

eldadHaber commented 6 years ago

Amazing work

On Feb 2, 2018, at 2:23 PM, DavidBegert notifications@github.com wrote:

30% faster for apply, 75% faster for JYmv, 55% faster for JYTmv 38% mem improvement for JYmv, 40% Mem improvement for JYTmv

You can view, comment on, or merge this pull request online at:

https://github.com/XtractOpen/Meganet.jl/pull/40

Commit Summary

batch norm optimized File Changes

M .gitignore (1) M benchmarks/micro/bm_batchnorm.jl (54) M src/layers/normLayer.jl (65) Patch Links:

https://github.com/XtractOpen/Meganet.jl/pull/40.patch https://github.com/XtractOpen/Meganet.jl/pull/40.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.