Skip to content

Instantly share code, notes, and snippets.

View ekinakyurek's full-sized avatar

Ekin Akyürek ekinakyurek

View GitHub Profile
@ekinakyurek
ekinakyurek / gan.ipynb
Last active October 15, 2018 03:24
GAN Notebook
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@ekinakyurek
ekinakyurek / resnet.jl
Last active December 19, 2018 09:43
Resnet Julia
#model url: https://github.com/FluxML/Metalhead.jl/releases/download/v0.1.1/resnet.bson
#install
using Knet, KnetLayers, BSON, Images
struct ResidualBlock
layers
shortcut
end
function ResidualBlock(filters, kernels::Array{Tuple{Int,Int}}, pads::Array{Tuple{Int,Int}}, strides::Array{Tuple{Int,Int}}, shortcut = identity)
@ekinakyurek
ekinakyurek / IndexedDict.jl
Last active January 25, 2019 13:48
IndexedDict
import Base: get, length, getindex, push!, append!
struct IndexedDict{T}
toIndex::Dict{T,Int};
toElement::Vector{T};
IndexedDict{T}(toIndex,toElement) where T = new(toIndex,toElement)
IndexedDict{T}(toIndex,toElement) where T<:Integer = error("Cannot Create IndexedDict of Integers")
IndexedDict{T}(toIndex,toElement) where T<:AbstractArray = error("Cannot Create IndexedDict of Arrays")
end
@ekinakyurek
ekinakyurek / benchmark.sh
Last active February 15, 2019 23:20
KnetArraysxCuarrays
#!/usr/bin/env bash
set -uxe
cd $(mktemp -d)
git clone https://github.com/ekinakyurek/Mac-Network .
sed -i '1d' benchmark.jl # remove pkg import/activate
export JULIA_DEPOT_PATH=$(mktemp -d)
julia --project -e 'using Pkg; Pkg.instantiate(); pkg"free Knet"; pkg"add Knet#cuarrays CuArrays#master CUDAnative#master CUDAdrv#master"'
@ekinakyurek
ekinakyurek / benchmark.log
Last active February 15, 2019 23:31
LogKnetArraysxCuarrays
[ec2-user@ip-172-31-14-239 ~]$ sh benchmark.sh
++ mktemp -d
+ cd /tmp/tmp.WbX6402Bko
+ git clone https://github.com/ekinakyurek/Mac-Network .
Cloning into '.'...
remote: Enumerating objects: 68, done.
remote: Counting objects: 100% (68/68), done.
remote: Compressing objects: 100% (52/52), done.
remote: Total 328 (delta 30), reused 47 (delta 15), pack-reused 260
Receiving objects: 100% (328/328), 29.89 MiB | 50.42 MiB/s, done.
@ekinakyurek
ekinakyurek / thescienceofdeeplearningcolloquia.md
Created March 15, 2019 19:06
Notes from The Science of Deep Learning Colloquia

DAY 1

1. The State of Deep Learning : Overview Talk (I) Amnon Shashua