Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
amazon-dsstne-2016-07
Search
ryo nakamaru
July 27, 2016
Programming
210
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
amazon-dsstne-2016-07
ryo nakamaru
July 27, 2016
More Decks by ryo nakamaru
See All by ryo nakamaru
AWSで楽をするサービスメッシュ入門/appmesh-trial
pottava
1
1.7k
reinforce-2019-recap-lt
pottava
2
4.2k
ScaleShift-jp-2019-summer
pottava
1
230
Firecracker とは何か/what is Firecracker
pottava
12
5.7k
ハイブリッド並列 on Kubernetes/hybrid-parallel-program-on-kubernetes
pottava
1
470
AWS Fargate + Code 兄弟で始める継続的デリバリー / Continuous Delivery with AWS Fargate and Code brothers
pottava
12
3.3k
Singularity と NVIDIA GPU Cloud で作る ハイブリッド機械学習環境の構築 / Building a hybrid environment for Machine Learning with Singularity and NGC
pottava
3
1.5k
明日から始めるちょい足し λ / get-started-with-aws-lambda
pottava
4
2.5k
NGC と Singularity によるハイブリッド機械学習環境 / A hybrid environment for Machine Learning with NGC and Singularity
pottava
0
540
Other Decks in Programming
See All in Programming
エンジニア向け会社紹介/Findy Company Profile
findyinc
6
360k
吝嗇家のためのAI活用 / AI development for miser - ChatGPT + Issue Driven Development
tooppoo
0
190
Generative UI & AI-Assistants for Your Angular Solutions
manfredsteyer
PRO
0
110
ソフトウェア設計に溶けるインフラ ― AWS CDK のインフラ認識論
konokenj
2
500
AIを活用したE2Eテスト実装効率化のあゆみ / ebisu-mobile-14-kotetu
kotetuco
0
170
どこまでゆるくて許されるのか
tk3fftk
0
490
Terraform標準の組織で AWS CDKをどう使うか
mu7889yoon
0
200
Performance Engineering for Everyone
elenatanasoiu
0
270
The Bowling Game- From Imperative to Functional Programming - Part 1
philipschwarz
PRO
0
330
アルゴリズムは何を圧縮しているのか ─ Haskell から育った「圧縮代数」というメンタルモデル
naoya
16
3.5k
Even G2とAWSで推しのエージェントを召喚しよう!
har1101
1
170
Haskell/Servantを通してWebミドルウェアを捉え直す
pizzacat83
1
580
Featured
See All Featured
Ten Tips & Tricks for a 🌱 transition
stuffmc
0
150
svc-hook: hooking system calls on ARM64 by binary rewriting
retrage
2
340
Making the Leap to Tech Lead
cromwellryan
135
10k
Testing 201, or: Great Expectations
jmmastey
46
8.2k
Site-Speed That Sticks
csswizardry
13
1.3k
Agile that works and the tools we love
rasmusluckow
331
22k
How Fast Is Fast Enough? [PerfNow 2025]
tammyeverts
3
650
Building a A Zero-Code AI SEO Workflow
portentint
PRO
0
630
Rebuilding a faster, lazier Slack
samanthasiow
85
9.6k
Optimizing for Happiness
mojombo
378
71k
How to make the Groovebox
asonas
2
2.3k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
37
6.5k
Transcript
DL for Recommendations OpenHyperScale ษڧձ 16.07.28
ͦͷલʹ 2
3 V1.0! http://mesos.apache.org/blog/mesos-1-0-0-released/ (Yesterday, Jul 27 2016)
4 GPU SUPPORT!! https://github.com/apache/mesos/blob/1.0.0/CHANGELOG
՝͋ͱগ͠ 5 https://issues.apache.org/jira/browse/MESOS-4424
6 WHAT ABOUT DC/OS? https://mesosphere.com/blog/2016/07/27/apache-mesos-reaches-1-0/
ଓ͖࣍ճ.. 7
ຊ 8
Ϩίϝϯυ 9
“͜ͷΛങͬͨਓ͜Μͳങ͍ͬͯ·͢” 10
11 http://aws.typepad.com/sajp/2016/07/generating-recommendations-at-amazon-scale-with-apache-spark-and-amazon-dsstne.html AMAZON ͷύʔιφϧϨίϝϯυΛ AWS Ͱ࣮
12 ֶशσʔλੜɿ Amazon EMR (cluster) + Apache Spark (CPU) ֶशɾਪɿ
ECS (cluster) + ECR (docker image) + Amazon DSSTNE (GPU) σʔλసૹɾڞ༗ɿ Amazon S3 Ϩίϝϯυੜڥͷߏ
13 ೖྗɾग़ྗϨΠϠʹ Amazon ͷेສΛஔ͠ɺ ݸਓ͝ͱʹ͜ͷϞσϧΛ༻͍ͯϨίϝϯυ͖͢Λਪଌɻ Ϩίϝϯσʔγϣϯੜ DNN Ϟσϧ ※ ࠨਤɺޡղΛট͘ֆͰ͢Έ·ͤΜɻ
࣮ࡍʹೖྗͱग़ྗಉ͡ϊʔυ
DSSTNE 14
15 εύʔεσʔλʹ࠷దԽ͠ɺͱεέʔϧΛॏࢹͨ͠ DNN with GPUs ͳϥΠϒϥϦɻAmazon ͕ OSS ͱͯ͠ެ։ɻ ࣗࣾͷ
DL ʹ·ͭΘΔ՝ղܾͷͨΊʹ࡞͞Εͨͷɻ AMAZON DSSTNE https://github.com/amznlabs/amazon-dsstne/
16 ɾͱͯ͑͠Δ͚ͩͷύϑΥʔϚϯεΛॏࢹ ɾϨίϝϯυੜࣗવݴޠཧղॲཧͰѻ͏εύʔεσʔλͰಛʹߴ ɹɹʢg2.8xlarge Πϯελϯε 1 Ͱ TensorFlow ͷ 2.1
ഒʣ ɾαʔό 1 × ෳ GPU Ͱͷԋࢉ͕͔ͳΓಘҙ ɹɹ-> GPU ͝ͱʹλεΫΛࣗಈࢄ ɹɹ-> GPU Λ૿ͤॲཧΛߴΊΒΕΔʢεέʔϧΞτʣ ɹɹɹ& ΑΓେ͖ͳϞσϧ͕ѻ͑Δ ɾωοτϫʔΫͷఆ͕ٛ Caffe ΑΓγϯϓϧ ɾΈࠐΈʹະରԠ ɾϦΧϨϯτݱࡏݶఆతαϙʔτɻ͍ͣΕϩʔυϚοϓʹ͋Δɻ ͦͷଞͷ DL ϥΠϒϥϦͱͷҧ͍ͳͲ https://github.com/amznlabs/amazon-dsstne/blob/master/FAQ.md