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
How we use GPUs in Cookpad
Search
Yuichiro Someya
November 06, 2017
Programming
0
130
How we use GPUs in Cookpad
@Tokyo Machine Learning Kitchen
https://tokyo-ml.github.io/
Yuichiro Someya
November 06, 2017
Tweet
Share
More Decks by Yuichiro Someya
See All by Yuichiro Someya
にんげんがさき 基盤はあと / Developers over ML platform
ayemos
0
13k
機械学習をスモールスタートさせる方法 / small machine learning
ayemos
3
2k
アットホームな分析基盤の作り方 / Homemade Machine Learning Toolkits
ayemos
1
950
サービス開発、機械学習、クラウド / the trinity of machine learning
ayemos
0
3.3k
成長を止めない機械学習のやり方 / Don't stop 'til you get enough (data).
ayemos
15
5.1k
AWS で加速する機械学習 / Accelerate Machine Learning with AWS
ayemos
1
310
クックパッドの機械学習基盤 2018 / Machine Learning Platform at Cookpad ~ 2018 ~
ayemos
15
19k
PyTorchとCaffe2とONNXと深層学習モデルのデプロイについて
ayemos
1
2.9k
クックパッドにおけるAWS GPUインスタンスの利用事例 / Powering by AWS GPU Instances in Cookpad Inc
ayemos
0
400
Other Decks in Programming
See All in Programming
AHC041解説
terryu16
0
330
PHPとAPI Platformで作る本格的なWeb APIアプリケーション(入門編) / phpcon 2024 Intro to API Platform
ttskch
0
390
混沌とした例外処理とエラー監視に秩序をもたらす
morihirok
13
2.1k
テストコード書いてみませんか?
onopon
2
330
週次リリースを実現するための グローバルアプリ開発
tera_ny
1
1.1k
PHPカンファレンス 2024|共創を加速するための若手の技術挑戦
weddingpark
0
130
見えないメモリを観測する: PHP 8.4 `pg_result_memory_size()` とSQL結果のメモリ管理
kentaroutakeda
0
930
Flatt Security XSS Challenge 解答・解説
flatt_security
0
710
VisionProで部屋の明るさを反映させるシェーダーを作った話
segur
0
100
ドメインイベント増えすぎ問題
h0r15h0
2
560
Findy Team+ Awardを受賞したかった!ベストプラクティス応募内容をふりかえり、開発生産性向上もふりかえる / Findy Team Plus Award BestPractice and DPE Retrospective 2024
honyanya
0
140
Androidアプリのモジュール分割における:x:commonを考える
okuzawats
1
270
Featured
See All Featured
Making the Leap to Tech Lead
cromwellryan
133
9k
BBQ
matthewcrist
85
9.4k
Large-scale JavaScript Application Architecture
addyosmani
510
110k
Rails Girls Zürich Keynote
gr2m
94
13k
Building an army of robots
kneath
302
45k
実際に使うSQLの書き方 徹底解説 / pgcon21j-tutorial
soudai
173
51k
Optimizing for Happiness
mojombo
376
70k
Imperfection Machines: The Place of Print at Facebook
scottboms
267
13k
Fontdeck: Realign not Redesign
paulrobertlloyd
82
5.3k
Distributed Sagas: A Protocol for Coordinating Microservices
caitiem20
330
21k
It's Worth the Effort
3n
183
28k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
507
140k
Transcript
)PXXFVTF(16TJO$PPLQBE :VJDIJSP4PNFZB!$PPLQBE*OD3%
‣ Yuichiro Someya (ayemos) ‣ github.com/ayemos ‣ Machine Learning Enginner
@ Cookpad Inc. # 2016(new grads) ~ Current
None
‣ 0VS(16FOWJSPONFOU )PXXFVUJMJ[F"84T(16JOTUBODFT )PXXFLFFQPVSTDBMBCJMJUZPGUFBNTJO3%
/7*%*"7
All-in on AWS since 2011
All-in on AWS since 2011 Amazon RDS (Relational Data)
Amazon Redshift (Data Warehouse)
All-in on AWS since 2011 Amazon S3 (Object Storage)
Amazon RDS (Relational Data) Amazon Redshift (Data Warehouse)
All-in on AWS since 2011 Amazon S3 (Object Storage)
Amazon RDS (Relational Data) Amazon Redshift (Data Warehouse) 7JSUVBM1SJWBUF$MPVE
7JSUVBM1SJWBUF$MPVE All-in on AWS since 2011 Amazon S3 (Object
Storage) Amazon RDS (Relational Data) Amazon Redshift (Data Warehouse) Amazon EC2 (Computation)
‣ $6%" ‣ DV%//
‣ $6%" ‣ DV%// (Snapshot)
‣ $6%" ‣ DV%// (Snapshot) ‣ $6%" ‣ DV%//
‣ $6%" ‣ DV%// (Snapshot) ‣ $6%" ‣ DV%//
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6
‣ $6%" ‣ DV%// (Snapshot) ‣ $6%" ‣ DV%//
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6
5FNQMBUF CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 5FNQMBUF KTPO
QBDLFSCVJME
‣ $6%" ‣ DV%// (Snapshot) ‣ $6%" ‣ DV%//
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6
‣ $6%" ‣ DV%// (Snapshot) ‣ $6%" ‣ DV%//
IUUQTBXTBNB[PODPNBNB[POBJBNJT CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... `ssh` ...
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... `ssh` ...
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... `ssh` ...
AWS Lambda (Function) Stop! Idle? (Hourly)
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... `ssh` ...
AWS Lambda (Function) Stop! Idle? (Hourly)
‣ 0OEFNBOE(16XPSLCFODIFT 6UJMJ[F".*UPNVMUJQMFXPSLCFODIFOWJSPONFOUT 1BDLFSNBLFTJUFBTJFSUPVQEBUFBOENPSFTUBCMF 0QFSBUFWJB$IBUCPU 8SBQVQ