際際滷

際際滷Share a Scribd company logo
Enhanced AI Platform
Produced by Tae Young Lee
NAS
Model
Model
Memory Limitation
Inference Speed
Worse Performance
Training Model
Model
Data
Data
Data
1)一危 覲 2)覈 覲
GPU
蠏 伎 螳 一危一  覈碁 れ企り
焔レ レ讌 
(  一危一 れ 覲伎  )
Serving Layer Model
旧  gradient 覈語 蠍一 觜襦蠍 覓語
覿 旧 牛  襯 襴朱,
覈語 貉れ 磯 旧 覲企 襷 螳 
Training Speed
覈 蠍郁 讀螳覃伎 豢襦 蟇碁Μ 螳 
企蠍 覓語 覓語螳 
覈語 貉れ覃伎 螳 襾殊 覓語螳
   蟆 覃覈襴 伎
Memory
Enhanced AI Platform
覲旧 GPU襯  旧 牛  覓語襯 願屋 
讌襷 覿 旧朱 覈語 牛朱  覓語
願屋讌 . ( GPU 蟲襷る NVidia襷 譬 (?) )
GPU
螳讌豺蠍 (Pruning)
螳譴豺 覿 (Weight Factorization)
讌 讀襯 (Knowledge Distillation)
螳譴豺 螻旧 (Weight Sharing)
 (Quantization)
Pre-train vs. Downstream
Model 豢 蠍一
Model Training
Data Labeling
Model Evaluation
Data Versioning
Model Service
Model Prediction
Model Deployment Model Versioning
Serving Architecture
Legacy Interface
Scaling Hardware
Model Life Cycle 譴
覈 覲 Training Image 覦
Workspace Training 
Training  Shared Memory れ
Training  Multi GPU れ
Model Serving Monitoring
Inference 螻 Gateway
POD Monitoring
Rancher 螳 れ襷朱 Node, Pod
襯 覈磯    蠍磯レ 螻
覲企 誤 襯 蠍 伎
Prometheus, Grafana 煙  覈磯
Data Selection Data Cleaning
Data Pre-Processing
Model MetaData 蟯襴
Model Validation
ML-Metadata
Model Registry
ML-Metadata Model Registry
Data Versioning
DW 一危一 蟲煙   一危 螻豸旧  襦 るジ 一危磯れ
危危  伎 覃 覈碁   螳螳 一危一れ  企レ
讌伎 覈 碁企  給 覈瑚骸 binding 蟯螻襯   蟆
る 襦 レ .
伎 覦 讌覲伎 蟯 り
 Data  覈語 狩 覦煙 蟯襴
https://deview.kr/2019/schedule/310
Training螻 Inference  Scaling Hardware
Training  蟲焔 Data Size Scope 磯 螳伎  Training
Resource谿願 譟伎
企ゼ 伎 Data Size Scope 磯 朱 Traing蟆 蟲煙 
IDE  覦 蟆 蟲煙 Resource Clustering 螻牛
Inference  朱 碁曙 れ伎 讌 覈磯殊 襷 GPU襯 覲危螻 
- Throughput 豢 覿螳 ( Resource 觜 )
Inference襯 蠍  豕 ル  覈襴 (?)
Scaling Hardware
https://deview.kr/2020/sessions/393
Model Validation
覈語 覲蟆渚  Inference蟆郁骸螳 伎蟆 る蟆 覦讌
Production(伎螻)螻 Staging(螳覦螻) 觜蟲蠍
https://deview.kr/2020/sessions/393
ML-metadata襯
Model Registry
覈瑚骸 覿螳 覲企ゼ 
- model-id, model-URI,
description, user, metrics 
れ 覈 殊 HDFS 
NAS 
https://deview.kr/2020/sessions/393
Serving蟯 れ 蟆 蟲煙 狩
Inference Speed Response Time 覲伎レ  Model Build  Base Image   朱 豌襴 磯ジ 蠍一 Base
Image Switching 牛 襴  
https://deview.kr/2019/schedule/310
覈 觜  襦語 蟲譟磯
https://deview.kr/2020/sessions/329
覦壱 螻  覦 企 蟯襴
CPU / GPU Cluster 覦壱  譯殊 企ろ ろ朱危 蟲
語 覦壱 蟲  
殊壱  誤  焔 / 蟯襴 蠍磯
襦 語ろ伎 覦壱
蠍一ヾ 語ろ伎 
EndPoint 覲
一危
 誤 れ擦覯襴
AI Platform 螻 螻ろ伎  
Scaling Hardware
Model Versioning
レ 讌 殊壱

More Related Content

What's hot (19)

[磯襭/141206 覦覩手啓] 豕 豐襴
[磯襭/141206 覦覩手啓]  豕  豐襴[磯襭/141206 覦覩手啓]  豕  豐襴
[磯襭/141206 覦覩手啓] 豕 豐襴
MinGeun Park
Cuda intro
Cuda introCuda intro
Cuda intro
譯殊
Ibm螻 nvidia螳 ル
Ibm螻 nvidia螳  ル Ibm螻 nvidia螳  ル
Ibm螻 nvidia螳 ル
ibmrep
OPEN_POWER8_SESSION_20150316
OPEN_POWER8_SESSION_20150316OPEN_POWER8_SESSION_20150316
OPEN_POWER8_SESSION_20150316
蠍壱 蟾
覈覦 蟆 豕
覈覦 蟆 豕 覈覦 蟆 豕
覈覦 蟆 豕
tartist
覲襦蠏碁覦螻 Cuda
覲襦蠏碁覦螻 Cuda覲襦蠏碁覦螻 Cuda
覲襦蠏碁覦螻 Cuda
Seok-joon Yun
Ndc2012 豕讌 れ 豢 蠍磯 螳
Ndc2012 豕讌 れ 豢 蠍磯 螳Ndc2012 豕讌 れ 豢 蠍磯 螳
Ndc2012 豕讌 れ 豢 蠍磯 螳
Jiho Choi
Compute shader DX11
Compute shader DX11Compute shader DX11
Compute shader DX11
覩殊
讌 貉危 伎 蟆 螳覦 蠍一 覦
讌 貉危 伎 蟆 螳覦 蠍一 覦讌 貉危 伎 蟆 螳覦 蠍一 覦
讌 貉危 伎 蟆 螳覦 蠍一 覦
Sukwoo Lee
Openface
OpenfaceOpenface
Openface
jaeho kang
願鍵譬 覃一 蠍磯 Open cv 襦 覦 危襴貅伎
願鍵譬 覃一 蠍磯 Open cv  襦 覦  危襴貅伎 願鍵譬 覃一 蠍磯 Open cv  襦 覦  危襴貅伎
願鍵譬 覃一 蠍磯 Open cv 襦 覦 危襴貅伎
Seunghwa Song
2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)
2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)
2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)
khuhacker
Compute shader
Compute shaderCompute shader
Compute shader
QooJuice
Data-Oriented Design螻 DOTS
Data-Oriented Design螻  DOTSData-Oriented Design螻  DOTS
Data-Oriented Design螻 DOTS
Sukwoo Lee
[264] large scale deep-learning_on_spark
[264] large scale deep-learning_on_spark[264] large scale deep-learning_on_spark
[264] large scale deep-learning_on_spark
NAVER D2
蟆襦語 GPGPU
蟆襦語  GPGPU蟆襦語  GPGPU
蟆襦語 GPGPU
YEONG-CHEON YOU
[NDC08] 豕 襦朱 - ′綾蠏
[NDC08] 豕 襦朱 - ′綾蠏[NDC08] 豕 襦朱 - ′綾蠏
[NDC08] 豕 襦朱 - ′綾蠏
ChangKyu Song
願鍵譬 覃一 襦語襯 襦蠏碁覦 語 覦 豌襴 ろ
願鍵譬 覃一 襦語襯  襦蠏碁覦 語 覦 豌襴 ろ願鍵譬 覃一 襦語襯  襦蠏碁覦 語 覦 豌襴 ろ
願鍵譬 覃一 襦語襯 襦蠏碁覦 語 覦 豌襴 ろ
Seunghwa Song
Chapter2 ap group11
Chapter2 ap group11Chapter2 ap group11
Chapter2 ap group11
Hyun Wong Choi
[磯襭/141206 覦覩手啓] 豕 豐襴
[磯襭/141206 覦覩手啓]  豕  豐襴[磯襭/141206 覦覩手啓]  豕  豐襴
[磯襭/141206 覦覩手啓] 豕 豐襴
MinGeun Park
Cuda intro
Cuda introCuda intro
Cuda intro
譯殊
Ibm螻 nvidia螳 ル
Ibm螻 nvidia螳  ル Ibm螻 nvidia螳  ル
Ibm螻 nvidia螳 ル
ibmrep
OPEN_POWER8_SESSION_20150316
OPEN_POWER8_SESSION_20150316OPEN_POWER8_SESSION_20150316
OPEN_POWER8_SESSION_20150316
蠍壱 蟾
覈覦 蟆 豕
覈覦 蟆 豕 覈覦 蟆 豕
覈覦 蟆 豕
tartist
覲襦蠏碁覦螻 Cuda
覲襦蠏碁覦螻 Cuda覲襦蠏碁覦螻 Cuda
覲襦蠏碁覦螻 Cuda
Seok-joon Yun
Ndc2012 豕讌 れ 豢 蠍磯 螳
Ndc2012 豕讌 れ 豢 蠍磯 螳Ndc2012 豕讌 れ 豢 蠍磯 螳
Ndc2012 豕讌 れ 豢 蠍磯 螳
Jiho Choi
Compute shader DX11
Compute shader DX11Compute shader DX11
Compute shader DX11
覩殊
讌 貉危 伎 蟆 螳覦 蠍一 覦
讌 貉危 伎 蟆 螳覦 蠍一 覦讌 貉危 伎 蟆 螳覦 蠍一 覦
讌 貉危 伎 蟆 螳覦 蠍一 覦
Sukwoo Lee
願鍵譬 覃一 蠍磯 Open cv 襦 覦 危襴貅伎
願鍵譬 覃一 蠍磯 Open cv  襦 覦  危襴貅伎 願鍵譬 覃一 蠍磯 Open cv  襦 覦  危襴貅伎
願鍵譬 覃一 蠍磯 Open cv 襦 覦 危襴貅伎
Seunghwa Song
2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)
2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)
2015 2 襴 伎擦 碁碁 - 覲貉危 螳 (16蠍 蟾)
khuhacker
Compute shader
Compute shaderCompute shader
Compute shader
QooJuice
Data-Oriented Design螻 DOTS
Data-Oriented Design螻  DOTSData-Oriented Design螻  DOTS
Data-Oriented Design螻 DOTS
Sukwoo Lee
[264] large scale deep-learning_on_spark
[264] large scale deep-learning_on_spark[264] large scale deep-learning_on_spark
[264] large scale deep-learning_on_spark
NAVER D2
[NDC08] 豕 襦朱 - ′綾蠏
[NDC08] 豕 襦朱 - ′綾蠏[NDC08] 豕 襦朱 - ′綾蠏
[NDC08] 豕 襦朱 - ′綾蠏
ChangKyu Song
願鍵譬 覃一 襦語襯 襦蠏碁覦 語 覦 豌襴 ろ
願鍵譬 覃一 襦語襯  襦蠏碁覦 語 覦 豌襴 ろ願鍵譬 覃一 襦語襯  襦蠏碁覦 語 覦 豌襴 ろ
願鍵譬 覃一 襦語襯 襦蠏碁覦 語 覦 豌襴 ろ
Seunghwa Song
Chapter2 ap group11
Chapter2 ap group11Chapter2 ap group11
Chapter2 ap group11
Hyun Wong Choi

Similar to Enhanced ai platform (20)

Infra as a model service
Infra as a model serviceInfra as a model service
Infra as a model service
Tae Young Lee
Create a solution including deep learning models
Create a solution including deep learning modelsCreate a solution including deep learning models
Create a solution including deep learning models
Tae Young Lee
Transfer learning usage
Transfer learning usageTransfer learning usage
Transfer learning usage
Tae Young Lee
蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓
蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓
蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓
cradon25
The Data tech for AI based innovation(蠍一 AI蠍磯 讌 一危 蠍一)
The Data tech for AI based innovation(蠍一 AI蠍磯  讌 一危 蠍一)The Data tech for AI based innovation(蠍一 AI蠍磯  讌 一危 蠍一)
The Data tech for AI based innovation(蠍一 AI蠍磯 讌 一危 蠍一)
SANG WON PARK
2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦
2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦
2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦
Suji Lee
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020 AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWSKRUG - AWS蟲覈
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
Jinwoong Kim
譬豢磯 Data Engineering 糾鍵
譬豢磯 Data Engineering 糾鍵譬豢磯 Data Engineering 糾鍵
譬豢磯 Data Engineering 糾鍵
DONGMIN LEE
ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)
ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)
ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)
Devgear
PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...
PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...
PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...
Sunghoon Joo
[AWS Innovate 殊 貉朱一] 覦焔 蠍郁 觜るゼ 觜覯 - れ谿, AWS 覦るΜろ
[AWS Innovate 殊 貉朱一] 覦焔   蠍郁  觜るゼ   觜覯 - れ谿, AWS  覦るΜろ[AWS Innovate 殊 貉朱一] 覦焔   蠍郁  觜るゼ   觜覯 - れ谿, AWS  覦るΜろ
[AWS Innovate 殊 貉朱一] 覦焔 蠍郁 觜るゼ 觜覯 - れ谿, AWS 覦るΜろ
Amazon Web Services Korea
AWS 觜 螻焔リ鍵螻(ML) 一企り鍵 - れ谿 覦るΜろ, AWS :: AWS Innovate 2019
AWS 觜  螻焔リ鍵螻(ML)  一企り鍵 - れ谿   覦るΜろ, AWS :: AWS Innovate 2019AWS 觜  螻焔リ鍵螻(ML)  一企り鍵 - れ谿   覦るΜろ, AWS :: AWS Innovate 2019
AWS 觜 螻焔リ鍵螻(ML) 一企り鍵 - れ谿 覦るΜろ, AWS :: AWS Innovate 2019
Amazon Web Services Korea
ル 瑚 覓誤蠍 覦 覿
ル 瑚 覓誤蠍 覦 覿ル 瑚 覓誤蠍 覦 覿
ル 瑚 覓誤蠍 覦 覿
Ubuntu Korea Community
蟠蠍壱_誤启Μ
蟠蠍壱_誤启Μ蟠蠍壱_誤启Μ
蟠蠍壱_誤启Μ
Kihoon4
襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″
襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″
襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″
Evion Kim
AWS CLOUD 2018- AWS 襦 牛 襾語 觜, Amazon SageMaker (蟾覓危 襭讀 ろ)
AWS CLOUD 2018- AWS 襦 牛 襾語  觜, Amazon SageMaker (蟾覓危 襭讀 ろ)AWS CLOUD 2018- AWS 襦 牛 襾語  觜, Amazon SageMaker (蟾覓危 襭讀 ろ)
AWS CLOUD 2018- AWS 襦 牛 襾語 觜, Amazon SageMaker (蟾覓危 襭讀 ろ)
Amazon Web Services Korea
Pycon korea 2018 kaggle tutorial(kaggle break)
Pycon korea 2018 kaggle tutorial(kaggle break)Pycon korea 2018 kaggle tutorial(kaggle break)
Pycon korea 2018 kaggle tutorial(kaggle break)
Yeonmin Kim
襾語螻 危結一 危
襾語螻 危結一 危襾語螻 危結一 危
襾語螻 危結一 危
豌覩 蟠
Microsoft loves communities - Korea DevRel Team
Microsoft loves communities - Korea DevRel TeamMicrosoft loves communities - Korea DevRel Team
Microsoft loves communities - Korea DevRel Team
Ian Choi
Infra as a model service
Infra as a model serviceInfra as a model service
Infra as a model service
Tae Young Lee
Create a solution including deep learning models
Create a solution including deep learning modelsCreate a solution including deep learning models
Create a solution including deep learning models
Tae Young Lee
Transfer learning usage
Transfer learning usageTransfer learning usage
Transfer learning usage
Tae Young Lee
蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓
蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓
蟆渚仰矧蟲喝雨壊温霞喝″_覈覃伎葎喝競_郁規喝豕脚覲願.沿糸韓
cradon25
The Data tech for AI based innovation(蠍一 AI蠍磯 讌 一危 蠍一)
The Data tech for AI based innovation(蠍一 AI蠍磯  讌 一危 蠍一)The Data tech for AI based innovation(蠍一 AI蠍磯  讌 一危 蠍一)
The Data tech for AI based innovation(蠍一 AI蠍磯 讌 一危 蠍一)
SANG WON PARK
2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦
2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦
2015 SINVAS USER CONFERENCE - MDD/MDA 螳覦覦覯襦 牛 覲伎ろ 螳覦覦
Suji Lee
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020 AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWSKRUG - AWS蟲覈
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
AWS蠍磯 覯襴 一危磯危 蟲豢蠍 - 蟾讌 (SK C&C) :: AWS Community Day 2020
Jinwoong Kim
譬豢磯 Data Engineering 糾鍵
譬豢磯 Data Engineering 糾鍵譬豢磯 Data Engineering 糾鍵
譬豢磯 Data Engineering 糾鍵
DONGMIN LEE
ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)
ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)
ER/Studio 一危 覈碁 襭朱 襷願係伎(from ERwin)
Devgear
PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...
PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...
PR-383: Solving ImageNet: a Unified Scheme for Training any Backbone to Top R...
Sunghoon Joo
[AWS Innovate 殊 貉朱一] 覦焔 蠍郁 觜るゼ 觜覯 - れ谿, AWS 覦るΜろ
[AWS Innovate 殊 貉朱一] 覦焔   蠍郁  觜るゼ   觜覯 - れ谿, AWS  覦るΜろ[AWS Innovate 殊 貉朱一] 覦焔   蠍郁  觜るゼ   觜覯 - れ谿, AWS  覦るΜろ
[AWS Innovate 殊 貉朱一] 覦焔 蠍郁 觜るゼ 觜覯 - れ谿, AWS 覦るΜろ
Amazon Web Services Korea
AWS 觜 螻焔リ鍵螻(ML) 一企り鍵 - れ谿 覦るΜろ, AWS :: AWS Innovate 2019
AWS 觜  螻焔リ鍵螻(ML)  一企り鍵 - れ谿   覦るΜろ, AWS :: AWS Innovate 2019AWS 觜  螻焔リ鍵螻(ML)  一企り鍵 - れ谿   覦るΜろ, AWS :: AWS Innovate 2019
AWS 觜 螻焔リ鍵螻(ML) 一企り鍵 - れ谿 覦るΜろ, AWS :: AWS Innovate 2019
Amazon Web Services Korea
蟠蠍壱_誤启Μ
蟠蠍壱_誤启Μ蟠蠍壱_誤启Μ
蟠蠍壱_誤启Μ
Kihoon4
襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″
襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″
襷語 Big Data Recommendation Products - 伎 一危磯ゼ 牛 伎殊 豸″
Evion Kim
AWS CLOUD 2018- AWS 襦 牛 襾語 觜, Amazon SageMaker (蟾覓危 襭讀 ろ)
AWS CLOUD 2018- AWS 襦 牛 襾語  觜, Amazon SageMaker (蟾覓危 襭讀 ろ)AWS CLOUD 2018- AWS 襦 牛 襾語  觜, Amazon SageMaker (蟾覓危 襭讀 ろ)
AWS CLOUD 2018- AWS 襦 牛 襾語 觜, Amazon SageMaker (蟾覓危 襭讀 ろ)
Amazon Web Services Korea
Pycon korea 2018 kaggle tutorial(kaggle break)
Pycon korea 2018 kaggle tutorial(kaggle break)Pycon korea 2018 kaggle tutorial(kaggle break)
Pycon korea 2018 kaggle tutorial(kaggle break)
Yeonmin Kim
襾語螻 危結一 危
襾語螻 危結一 危襾語螻 危結一 危
襾語螻 危結一 危
豌覩 蟠
Microsoft loves communities - Korea DevRel Team
Microsoft loves communities - Korea DevRel TeamMicrosoft loves communities - Korea DevRel Team
Microsoft loves communities - Korea DevRel Team
Ian Choi

More from Tae Young Lee (20)

DeepSeek梶 釈 Trend (Faculty Tae Young Lee)
DeepSeek梶 釈  Trend (Faculty Tae Young Lee)DeepSeek梶 釈  Trend (Faculty Tae Young Lee)
DeepSeek梶 釈 Trend (Faculty Tae Young Lee)
Tae Young Lee
Transitioning from the Era of Big Data to LLMs_Deriving Insights
Transitioning from the Era of Big Data to LLMs_Deriving InsightsTransitioning from the Era of Big Data to LLMs_Deriving Insights
Transitioning from the Era of Big Data to LLMs_Deriving Insights
Tae Young Lee
Facebook Meta's technical direction in Large Language Models (LLMs)
Facebook Meta's technical direction in Large Language Models (LLMs)Facebook Meta's technical direction in Large Language Models (LLMs)
Facebook Meta's technical direction in Large Language Models (LLMs)
Tae Young Lee
MultiModal Embedding integrates various data types, like images, text, and au...
MultiModal Embedding integrates various data types, like images, text, and au...MultiModal Embedding integrates various data types, like images, text, and au...
MultiModal Embedding integrates various data types, like images, text, and au...
Tae Young Lee
Review of the Paper on Capabilities of Gemini Models in Medicine
Review of the Paper on Capabilities of Gemini Models in MedicineReview of the Paper on Capabilities of Gemini Models in Medicine
Review of the Paper on Capabilities of Gemini Models in Medicine
Tae Young Lee
A future that integrates LLMs and LAMs (Symposium)
A future that integrates LLMs and LAMs (Symposium)A future that integrates LLMs and LAMs (Symposium)
A future that integrates LLMs and LAMs (Symposium)
Tae Young Lee
LLMs Service that provides what users want to know
LLMs Service that provides what users want to knowLLMs Service that provides what users want to know
LLMs Service that provides what users want to know
Tae Young Lee
Exploring Deep Learning Acceleration Technology Embedded in LLMs
Exploring Deep Learning Acceleration Technology Embedded in LLMsExploring Deep Learning Acceleration Technology Embedded in LLMs
Exploring Deep Learning Acceleration Technology Embedded in LLMs
Tae Young Lee
LLM 求= メ求 求メ =求
LLM  求= メ求 求メ =求LLM  求= メ求 求メ =求
LLM 求= メ求 求メ =求
Tae Young Lee
Recommendation System History
Recommendation System HistoryRecommendation System History
Recommendation System History
Tae Young Lee
蟆讌 ChatGPT
蟆讌  ChatGPT蟆讌  ChatGPT
蟆讌 ChatGPT
Tae Young Lee
メ戟≡求 メ釈 求午メ求 =釈メ午
メ戟≡求 メ釈 求午メ求  =釈メ午メ戟≡求 メ釈 求午メ求  =釈メ午
メ戟≡求 メ釈 求午メ求 =釈メ午
Tae Young Lee
PaLM Paper Review
PaLM Paper ReviewPaLM Paper Review
PaLM Paper Review
Tae Young Lee
The Deep Learning Compiler
The Deep Learning CompilerThe Deep Learning Compiler
The Deep Learning Compiler
Tae Young Lee
History of Vision AI
History of Vision AIHistory of Vision AI
History of Vision AI
Tae Young Lee
Aws architecture
Aws architectureAws architecture
Aws architecture
Tae Young Lee
Real estate trust solution
Real estate trust solutionReal estate trust solution
Real estate trust solution
Tae Young Lee
Bayesian learning
Bayesian learningBayesian learning
Bayesian learning
Tae Young Lee
Deep learning financial product recommendation model
Deep learning financial product recommendation modelDeep learning financial product recommendation model
Deep learning financial product recommendation model
Tae Young Lee
Deep Learning Through Various Processes
Deep Learning Through Various ProcessesDeep Learning Through Various Processes
Deep Learning Through Various Processes
Tae Young Lee
DeepSeek梶 釈 Trend (Faculty Tae Young Lee)
DeepSeek梶 釈  Trend (Faculty Tae Young Lee)DeepSeek梶 釈  Trend (Faculty Tae Young Lee)
DeepSeek梶 釈 Trend (Faculty Tae Young Lee)
Tae Young Lee
Transitioning from the Era of Big Data to LLMs_Deriving Insights
Transitioning from the Era of Big Data to LLMs_Deriving InsightsTransitioning from the Era of Big Data to LLMs_Deriving Insights
Transitioning from the Era of Big Data to LLMs_Deriving Insights
Tae Young Lee
Facebook Meta's technical direction in Large Language Models (LLMs)
Facebook Meta's technical direction in Large Language Models (LLMs)Facebook Meta's technical direction in Large Language Models (LLMs)
Facebook Meta's technical direction in Large Language Models (LLMs)
Tae Young Lee
MultiModal Embedding integrates various data types, like images, text, and au...
MultiModal Embedding integrates various data types, like images, text, and au...MultiModal Embedding integrates various data types, like images, text, and au...
MultiModal Embedding integrates various data types, like images, text, and au...
Tae Young Lee
Review of the Paper on Capabilities of Gemini Models in Medicine
Review of the Paper on Capabilities of Gemini Models in MedicineReview of the Paper on Capabilities of Gemini Models in Medicine
Review of the Paper on Capabilities of Gemini Models in Medicine
Tae Young Lee
A future that integrates LLMs and LAMs (Symposium)
A future that integrates LLMs and LAMs (Symposium)A future that integrates LLMs and LAMs (Symposium)
A future that integrates LLMs and LAMs (Symposium)
Tae Young Lee
LLMs Service that provides what users want to know
LLMs Service that provides what users want to knowLLMs Service that provides what users want to know
LLMs Service that provides what users want to know
Tae Young Lee
Exploring Deep Learning Acceleration Technology Embedded in LLMs
Exploring Deep Learning Acceleration Technology Embedded in LLMsExploring Deep Learning Acceleration Technology Embedded in LLMs
Exploring Deep Learning Acceleration Technology Embedded in LLMs
Tae Young Lee
LLM 求= メ求 求メ =求
LLM  求= メ求 求メ =求LLM  求= メ求 求メ =求
LLM 求= メ求 求メ =求
Tae Young Lee
Recommendation System History
Recommendation System HistoryRecommendation System History
Recommendation System History
Tae Young Lee
メ戟≡求 メ釈 求午メ求 =釈メ午
メ戟≡求 メ釈 求午メ求  =釈メ午メ戟≡求 メ釈 求午メ求  =釈メ午
メ戟≡求 メ釈 求午メ求 =釈メ午
Tae Young Lee
PaLM Paper Review
PaLM Paper ReviewPaLM Paper Review
PaLM Paper Review
Tae Young Lee
The Deep Learning Compiler
The Deep Learning CompilerThe Deep Learning Compiler
The Deep Learning Compiler
Tae Young Lee
History of Vision AI
History of Vision AIHistory of Vision AI
History of Vision AI
Tae Young Lee
Real estate trust solution
Real estate trust solutionReal estate trust solution
Real estate trust solution
Tae Young Lee
Bayesian learning
Bayesian learningBayesian learning
Bayesian learning
Tae Young Lee
Deep learning financial product recommendation model
Deep learning financial product recommendation modelDeep learning financial product recommendation model
Deep learning financial product recommendation model
Tae Young Lee
Deep Learning Through Various Processes
Deep Learning Through Various ProcessesDeep Learning Through Various Processes
Deep Learning Through Various Processes
Tae Young Lee

Enhanced ai platform

  • 1. Enhanced AI Platform Produced by Tae Young Lee
  • 2. NAS Model Model Memory Limitation Inference Speed Worse Performance Training Model Model Data Data Data 1)一危 覲 2)覈 覲 GPU 蠏 伎 螳 一危一 覈碁 れ企り 焔レ レ讌 ( 一危一 れ 覲伎 ) Serving Layer Model 旧 gradient 覈語 蠍一 觜襦蠍 覓語 覿 旧 牛 襯 襴朱, 覈語 貉れ 磯 旧 覲企 襷 螳 Training Speed 覈 蠍郁 讀螳覃伎 豢襦 蟇碁Μ 螳 企蠍 覓語 覓語螳 覈語 貉れ覃伎 螳 襾殊 覓語螳 蟆 覃覈襴 伎 Memory Enhanced AI Platform 覲旧 GPU襯 旧 牛 覓語襯 願屋 讌襷 覿 旧朱 覈語 牛朱 覓語 願屋讌 . ( GPU 蟲襷る NVidia襷 譬 (?) ) GPU 螳讌豺蠍 (Pruning) 螳譴豺 覿 (Weight Factorization) 讌 讀襯 (Knowledge Distillation) 螳譴豺 螻旧 (Weight Sharing) (Quantization) Pre-train vs. Downstream Model 豢 蠍一
  • 3. Model Training Data Labeling Model Evaluation Data Versioning Model Service Model Prediction Model Deployment Model Versioning Serving Architecture Legacy Interface Scaling Hardware Model Life Cycle 譴 覈 覲 Training Image 覦 Workspace Training Training Shared Memory れ Training Multi GPU れ Model Serving Monitoring Inference 螻 Gateway POD Monitoring Rancher 螳 れ襷朱 Node, Pod 襯 覈磯 蠍磯レ 螻 覲企 誤 襯 蠍 伎 Prometheus, Grafana 煙 覈磯 Data Selection Data Cleaning Data Pre-Processing Model MetaData 蟯襴 Model Validation ML-Metadata Model Registry ML-Metadata Model Registry
  • 4. Data Versioning DW 一危一 蟲煙 一危 螻豸旧 襦 るジ 一危磯れ 危危 伎 覃 覈碁 螳螳 一危一れ 企レ 讌伎 覈 碁企 給 覈瑚骸 binding 蟯螻襯 蟆 る 襦 レ . 伎 覦 讌覲伎 蟯 り
  • 5. Data 覈語 狩 覦煙 蟯襴 https://deview.kr/2019/schedule/310
  • 6. Training螻 Inference Scaling Hardware Training 蟲焔 Data Size Scope 磯 螳伎 Training Resource谿願 譟伎 企ゼ 伎 Data Size Scope 磯 朱 Traing蟆 蟲煙 IDE 覦 蟆 蟲煙 Resource Clustering 螻牛 Inference 朱 碁曙 れ伎 讌 覈磯殊 襷 GPU襯 覲危螻 - Throughput 豢 覿螳 ( Resource 觜 ) Inference襯 蠍 豕 ル 覈襴 (?)
  • 8. Model Validation 覈語 覲蟆渚 Inference蟆郁骸螳 伎蟆 る蟆 覦讌 Production(伎螻)螻 Staging(螳覦螻) 觜蟲蠍 https://deview.kr/2020/sessions/393
  • 10. Model Registry 覈瑚骸 覿螳 覲企ゼ - model-id, model-URI, description, user, metrics れ 覈 殊 HDFS NAS https://deview.kr/2020/sessions/393
  • 11. Serving蟯 れ 蟆 蟲煙 狩 Inference Speed Response Time 覲伎レ Model Build Base Image 朱 豌襴 磯ジ 蠍一 Base Image Switching 牛 襴 https://deview.kr/2019/schedule/310
  • 12. 覈 觜 襦語 蟲譟磯 https://deview.kr/2020/sessions/329
  • 13. 覦壱 螻 覦 企 蟯襴 CPU / GPU Cluster 覦壱 譯殊 企ろ ろ朱危 蟲 語 覦壱 蟲 殊壱 誤 焔 / 蟯襴 蠍磯 襦 語ろ伎 覦壱 蠍一ヾ 語ろ伎 EndPoint 覲 一危 誤 れ擦覯襴 AI Platform 螻 螻ろ伎 Scaling Hardware Model Versioning レ 讌 殊壱