$30 off During Our Annual Pro Sale. View Details »
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Vertex AI 試してみた / tried-vertex-ai
Search
kasacchiful
April 24, 2022
Programming
0
430
Vertex AI 試してみた / tried-vertex-ai
2022/04/24 (日) Python機械学習勉強会 in 新潟で発表した資料です。
kasacchiful
April 24, 2022
Tweet
Share
More Decks by kasacchiful
See All by kasacchiful
dbt coreとFargateでデータ変換 / 20240928-jawsug-toyama-hokuriku-shinkansen
kasacchiful
1
51
What we keep in mind when migrating from Serverless Framework to AWS CDK and AWS SAM
kasacchiful
1
260
AWSでIcebergを使ってデータウェアハウスを構築してみる / 20240810-jawsug-akita
kasacchiful
0
31
サーバーレスパターンを元にAWS CDKでデータ基盤を構築する / 20240731_classmethod_odyssey_online_build_a_data_infrastructures_using_aws_cdk_based_on_serverless_patterns
kasacchiful
0
410
AWS IoT 1-clickがサービス終了するので、SORACOMに移行した話 / 20240518-jawsug-niigata-iotlt-niigata
kasacchiful
0
220
AWS Application Composerで始める、 サーバーレスなデータ基盤構築 / 20240406-jawsug-hokuriku-shinkansen
kasacchiful
1
530
AWSの各種サービス紹介と活用方法 − AI・ML活用デモを交えて − / 20231208aws-aiml-seminar
kasacchiful
0
500
Amazon Rekognition デモ / 20231208-aws_seminar-01-rekognition-demo
kasacchiful
0
520
Amazon Lookout for Vision デモ / 20231208-aws_seminar-02-lookout-vision-demo
kasacchiful
0
510
Other Decks in Programming
See All in Programming
HTTP compression in PHP and Symfony apps
dunglas
2
1.5k
テスト自動化失敗から再挑戦しチームにオーナーシップを委譲した話/STAC2024 macho
ma_cho29
1
860
Modular Monolith Monorepo ~シンプルさを保ちながらmonorepoのメリットを最大化する~
yuisakamoto
11
4k
急成長期の品質とスピードを両立するフロントエンド技術基盤
soarteclab
0
790
romajip: 日本の住所CSVデータを活用した英語住所変換ライブラリを作った話
sangunkang
0
2.4k
コンテンツの主権を守るため(?)、高機能画像CDNからAWS自前対応に乗り換えた話
lengthtail
1
120
Djangoの開発環境で工夫したこと - pre-commit / DevContainer
hiroki_yod
1
670
flutterkaigi_2024.pdf
kyoheig3
0
470
Refactor your code - refactor yourself
xosofox
1
170
Java 23の概要とJava Web Frameworkの現状 / Java 23 and Java web framework
kishida
2
380
Chrome DevTools: State of the Union 2024 - Debugging React & Beyond
addyosmani
1
130
DevFest Tokyo 2025 - Flutter のアプリアーキテクチャ現在地点
wasabeef
4
620
Featured
See All Featured
Understanding Cognitive Biases in Performance Measurement
bluesmoon
26
1.4k
Optimizing for Happiness
mojombo
376
70k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
29
2.3k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
132
33k
Code Review Best Practice
trishagee
64
17k
How To Stay Up To Date on Web Technology
chriscoyier
789
250k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
365
25k
Agile that works and the tools we love
rasmusluckow
328
21k
Become a Pro
speakerdeck
PRO
25
5k
Done Done
chrislema
181
16k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
44
9.3k
The Power of CSS Pseudo Elements
geoffreycrofte
73
5.4k
Transcript
Vertex AI ࢼͯ͠Έͨ Python ػցֶशษڧձ in ৽ׁ 2022-04-24 @kasacchiful
Classmethod, Inc. Solutions Architect / Software Develper Favorite: Community: •
JAWS-UG Niigata • Python ML in Niigata • JaSST Niigata • ASTER • SWANII • etc. Hiroshi Kasahara @kasacchiful @kasacchiful 2
MLOpsͱԿ͔ʁ
MLOps • ػցֶशϓϩδΣΫτΛԁʹਐΊΔͨΊͷऔΓΈɾମ੍ɾࢥ • ػցֶशνʔϜ / ։ൃνʔϜ / ӡ༻νʔϜ ͕͓ޓ͍ʹڠௐ͠߹͏͜ͱ
ͰɺػցֶशϞσϧͷ࣮͔Βӡ༻·ͰͷϥΠϑαΠΫϧΛԁʹਐ ΊɺܧଓతʹՁΛఏڙ͢Δ͜ͱΛతʹͯ͠Δ • DevOpsͷػցֶशϓϩδΣΫτ൛ • ʮ։ൃαΠΫϧͷॖʯʮ։ൃࣗମͷਝԽʯʮ৴པੑߴ͍ϦϦʔεʯ ΛMLϓϩδΣΫτʹ
MLγεςϜͷཁૉ IUUQTDMPVEHPPHMFDPNBSDIJUFDUVSFNMPQTDPOUJOVPVTEFMJWFSZBOEBVUPNBUJPOQJQFMJOFTJONBDIJOFMFBSOJOH
MLOps Life Cycle IUUQTOFBMBOBMZUJDTDPNFYQFSUJTFNMPQT
Continuous Delivery for Machine Learning end- to-end Process IUUQTNBSUJOGPXMFSDPNBSUJDMFTDENMIUNM
MLOpsΠϯϑϥج൫ʹඞཁͳ͜ͱ • σʔλιʔεͱɺ͔ͦ͜Βੜ͞ΕΔσʔληοτͷཧ • ֶशࡁϞσϧͷཧ • CI / CDؚΜͩMLσϦόϦύΠϓϥΠϯ •
Ұ࿈ͷδϣϒΛ؆୯ʹճͨ͢Ίͷίϯςφ
Ұ͔Βߏங͢ΔͷେมͳͷͰ ΫϥυαʔϏε͍͍ͨ
MLOpsؔ࿈αʔϏε • AWS • Amazon SageMaker • Google Cloud •
Vertex AI • Azure • Azure Machine Learning
IUUQTQBHFTBXTDMPVEDPNST5;.JNBHFT@"84Ͱߏங͢Δ.-0QTج൫@൛@@"*.-5PLZPQEG
IUUQTDMPVEHPPHMFDPNCMPHKBUPQJDTEFWFMPQFSTQSBDUJUJPOFSTOFXNMMFBSOJOHQBUIWFSUFYBJ
ࠓճVertex AIΛࢼͯ͠Έͨ
Vertex AIࢼͯ͠Έͨ • ϞσϧͷτϨʔχϯάɺσϓϩΠɺςετΛҰ௨Γࢼ͢ • AutoMLͷΫΠοΫελʔτΛϕʔεʹ࣮ࢪ • https://cloud.google.com/automl-tables/docs/quickstart?hl=ja • දܗࣜσʔλͷྨ
σʔληοτ
σʔληοτ
σʔληοτ
τϨʔχϯά
τϨʔχϯά
τϨʔχϯά
τϨʔχϯά
σϓϩΠ
σϓϩΠ
σϓϩΠ
σϓϩΠ
ςετ
curlͰࢼ͢ curl -X POST \ -H "Authorization: Bearer $(gcloud auth
print-access-token)" \ -H "Content-Type: application/json" \ https://asia-northeast1-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/asia-northeast1/endpoints/${ENDPOINT_ID}:predict \ -d “@${INPUT_DATA_FILE} " { "predictions": [ { "classes": [ "1" , "2 " ] , "scores": [ 0.98835468292236328 , 0.0116453049704432 5 ] } ] , "deployedModelId": "8383556259466444800" , "model": "projects/1066851579090/locations/asia-northeast1/models/6083061274810253312" , "modelDisplayName": "sample_tabular_dataset_model " }
ύΠϓϥΠϯ • kube fl ow pipelines ͘͠ Tensor fl ow
Extended ͰύΠϓϥ ΠϯΛهࡌͯ͠ɺVertex AIʹొ IUUQTDMPVEHPPHMFDPNCMPHKBQSPEVDUTBJNBDIJOFMFBSOJOHTFSWFSMFTTNBDIJOFMFBSOJOHQJQFMJOFTPOHPPHMFDMPVE
MLOpsΠϯϑϥج൫ʹඞཁͳ͜ͱ • σʔλιʔεͱɺ͔ͦ͜Βੜ͞ΕΔσʔληοτͷཧ • Vertex AIͰ֤छσʔληοτΛཧͰ͖Δ • ֶशࡁϞσϧͷཧ • Ϟσϧͷόʔδϣϯͷཧ͕Մೳ
• CI / CDؚΜͩMLσϦόϦύΠϓϥΠϯ • kebe fl ow pipelines͘͠Tensor fl ow ExtendedͰߏஙՄೳ • Ұ࿈ͷδϣϒΛ؆୯ʹճͨ͢Ίͷίϯςφ • طଘͷίϯςφར༻ͷ΄͔ɺΧελϜίϯςφ༻Մೳ
SageMaker / Vertex AI ͷػೳ σʔληοτ ͷ࡞ ਓྗʹΑΔΞ ϊςʔγϣϯ ࢧԉ
ಛྔϦϙδ τϦ ύΠϓϥΠϯ ϞσϧͷධՁ σϓϩΠ ϞχλϦϯά 4BHF.BLFS 4BHF.BLFS %BUB 8SBOHMFS 4BHF.BLFS (SPVOE5SVU I 4BHF.BLFS 'FBUVSF 4UPSF 4BHF.BLFS 1JQFMJOF 4BHF.BLFS %FCVHHFS 4BHF.BLFS $MBSJGZ όονਪ˓ ϦΞϧλΠϜ ਪ˓ 4BHF.BLFS .PEFM .POJUFS 7FSUFY"* σʔληοτ ϥϕϧλε Ϋ 7FSUFY"* 'FBUVSF 4UPSF 7FSUFY"* 1JQFMJOF ϞσϧͷධՁ όονਪ˓ ϦΞϧλΠϜ ਪ˓ 7FSUFY"* .POJUPSJOH
ॴײ • ݁ߏGUI্Ͱૢ࡞Ͱ͖Δ͠ɺΘ͔Γ͍͢ • ಛʹAutoML • Vertex AIͷPython SDK͕༻ҙ͞ΕͯΔͷͰɺίʔυͰཧͰ͖Δ •
SageMakerPython SDK͕͋Δ • ଞͷGoogle CloudαʔϏεͱͷ࿈ܞํ๏͕Α͘Θ͔Βͳ͍ (ଟͰ͖Δ ΜͩΖ͏͚Ͳ)
·ͱΊ • MLγεςϜߏஙͷࡍɺMLίʔυҎ֎ͷཁૉଟ͘ɺ࣌ؒखؒ ͔͔Δɻ • ΫϥυαʔϏεͷMLOpsؔ࿈ͷαʔϏεΛ͏·͘͏͜ͱͰɺ࿑ ྗΛগͳͯ͘͠MLγεςϜΛߏஙɾӡ༻͢Δ͜ͱ͕Մೳ
͓͠·͍