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
Google ColabでDL入門#2
Search
masa-ita
October 13, 2018
Technology
2
250
Google ColabでDL入門#2
Google Colab上でKeras Tutorialsの日本語版を試すハンズオンの2回め。
Python機械学習勉強会in新潟 2018-10-13での発表スライド。
masa-ita
October 13, 2018
Tweet
Share
More Decks by masa-ita
See All by masa-ita
Run Instant NeRF on Docker
itagakim
1
1.9k
3D Clustering and Metric Learning
itagakim
0
230
Cloud TPUの使い方〜BigBirdの日本語学習済みモデルを作る〜
itagakim
0
520
多言語学習済みモデルmT5とは?
itagakim
1
440
AWSのGPUを安く使ってTensorFlowモデルを訓練する方法
itagakim
0
260
最近の自然言語処理モデルの動向
itagakim
1
480
ディープラーニングで芸術はできるか?〜生成系ネットワークの進展〜
itagakim
0
260
AWSとTerraform初心者がやってみたこと
itagakim
1
360
IntroductionToTensorFlow2_0.pdf
itagakim
1
280
Other Decks in Technology
See All in Technology
Oracle Cloud Infrastructure:2024年4月度サービス・アップデート
oracle4engineer
PRO
1
110
「手動オペレーションに定評がある」と言われた私が心がけていること / phpcon_odawara2024
blue_goheimochi
2
320
[PlatformCon 24] Platform Orchestrators: The Missing Middle of Internal Developer Platforms?
danielbryantuk
1
180
カオナビの利用実績をアウトカムへつなげる旅 / example-of-data-management-startup-in-kaonavi
kaonavi
0
120
0→1開発における技術選定において一番大切なこと
bicstone
1
330
Databricksを活用してDELISH KITCHENのレシピレコメンドを開発した話
furu8
0
250
Databricks における 『MLOps』
databricksjapan
2
140
Tableau事例紹介 / Tableau Case Study of Eureka
kazuya_araki_tokyo
1
170
Databricks におけるデータエンジニアリング
databricksjapan
0
380
Janus
bkuhlmann
1
490
KubeCon EU 2024 Recap “Kubernetes Policy Time Machine: Where to Next?”
ryysud
0
120
自動生成を活用した、運用保守コストを抑える Error/Alert/Runbook の一元集約管理 / Centralized management of Error/Alert/Runbook to minimize operational costs using automated code generation
biwashi
9
2.1k
Featured
See All Featured
Debugging Ruby Performance
tmm1
70
11k
Building a Modern Day E-commerce SEO Strategy
aleyda
16
6.4k
Reflections from 52 weeks, 52 projects
jeffersonlam
344
19k
Bash Introduction
62gerente
604
210k
CSS Pre-Processors: Stylus, Less & Sass
bermonpainter
352
28k
Ruby is Unlike a Banana
tanoku
96
10k
4 Signs Your Business is Dying
shpigford
175
21k
Pencils Down: Stop Designing & Start Developing
hursman
117
11k
Build your cross-platform service in a week with App Engine
jlugia
225
17k
A better future with KSS
kneath
231
16k
Why Our Code Smells
bkeepers
PRO
331
56k
Testing 201, or: Great Expectations
jmmastey
27
6.3k
Transcript
Pythonػցֶशษڧձ in ৽ׁ ͰDLೖ#2 Google Colab 1 ൘֞ ਖ਼හ 2018-10-13
PyML in Niigata • ൘֞ ਖ਼හ • גࣜձࣾBSNΞΠωοτ ٕज़ސ •
தখاۀஅ࢜ • ຊRubyͷձ • ৽ׁΦʔϓϯιʔεڠձ • Pythonॳ৺ऀ WHO AM I ? 2
PyML in Niigata • ʮColaboratory ɺػց ֶशͷڭҭݚڀͷଅਐ Λతͱͨ͠ Google ݚ
ڀϓϩδΣΫτͰ͢ʯby Google • Google Driveͱ࿈ಈ • ίϯςφٕज़Λ׆༻ͨ͠ JupyterͷΧελϜڥ • νʔϜϝϯόʔͰͷڞಉ ฤू͕Մೳ • GPU/TPUແྉͰ͑ Δʂ • ੍ݶ࣌ؒ12࣌ؒʁ WHAT IS GOOGLE COLABORATORY? 3
ʮDriveʹίϐʔʯΛΫϦοΫ ͯ͠Driveͱͷ࿈ܞΛ։࢝ PyML in Niigata HOW TO BEGIN https://colab.research.google.com/ 4
PyML in Niigata • TensorFlowGoogleͷ OSSʢOpen Source Softwareʣ • Deep
LearningͷCoreͰ͋ ΔTensorԋࢉΛCPU/GPU Ͱߴ࣮ߦ͢ΔͨΊͷϥΠ ϒϥϦ • cf Caffe, MXNet, CNTK etc. • KerasFrançois Chollé @Google͕։ൃͨ͠OSS • TensorFlowɺ Theanoɺ CNTKͳͲͷόοΫΤϯυ Λ͍ɺModelΛॻ͖͢ ͘͢ΔϥΠϒϥϦ • TensorFlowʹࠐࡁ • cf Chainer, PyTorch etc. WHAT IS TENSORFLOW/KERAS? 5
ΦϦδφϧͷTutorials ɺࠨهͷURLͰެ։ ͞Ε͍ͯ·͢ɻ ͜ͷTutorials TensorFlowͷυΩϡϝ ϯτதͷhttps:// github.com/tensorflow/ docs/tree/master/site/ en/tutorials/keras ʹؚ
·Ε͍ͯ·͢ɻ PyML in Niigata LET’S START KERAS TUTORIALS https://www.tensorflow.org/tutorials/ 6
github.comͰ্هͷ ιʔείʔυʢJupyter NotebookʣΛදࣔ͠· ͢ɻ ࠨਤͷͱ͓Γɺ”Run in Google Colab (Japanese translation)”
ͷϦϯΫΛΫϦοΫ͢ ΔͱɺGoogle ColabͰ ։͘͜ͱ͕ग़དྷ·͢ɻ PyML in Niigata LET’S START KERAS WITH JAPANESE https://github.com/masa-ita/keras-tutorials/ 7
GITHUBλϒΛબ ͠ɺϢʔβʔ໊”masa- ita”Λೖྗͯ͠ݕࡧϘλ ϯΛΫϦοΫ͠·͢ɻ ϦϙδτϦ͔Β”masa- ita/keras-tutorials”ɺϒ ϥϯν”master”Λબ ͠·͢ɻ දࣔ͞Εͨύεͷத͔ Β”basic_text_classific
ation.ipynb”ͷӈͷϘ λϯΛΫϦοΫ͠·͢ɻ PyML in Niigata LET’S START KERAS WITH JAPANESE https://colab.research.google.com/ Λ։͖·͢ɻ 8
͜ͷঢ়ଶͰɺ Notebookͷ࣮ߦग़དྷ ·͕͢ɺग़ྗΛอଘ͢ Δ͜ͱ͕ग़དྷ·ͤΜɻ ग़ྗΛอଘ͢Δʹ ʮϑΝΠϧʯϝχϡʔ ͔ΒʮυϥΠϒʹίϐʔ ΛอଘʯΛબͯ͠ɺ Google Driveʹίϐʔ
Λอଘ͠·͢ɻ PyML in Niigata SAVE THE NOTEBOOK ON GOOGLE DRIVE 9
GPUΛ༻͢ΔʹɺʮϥϯλΠ Ϝʯϝχϡʔ͔ΒʮϥϯλΠϜͷλ ΠϓΛมߋʯΛબ͠ɺʮϋʔυΣ ΞΞΫηϥϨʔλʯΛʮNoneʯ͔ ΒʮGPUʯʹมߋ͠·͢ɻ PyML in Niigata HOW TO
USE GPU 10
Google͕࡞ͨ͠ϊʔτ ϒοΫҎ֎Λ࣮ߦ͢Δࡍ ʹɺࠨਤͷΑ͏ͳηΩϡ ϦςΟܯࠂ͕දࣔ͞ΕΔ ͜ͱ͕͋Γ·͢ɻ ϦηοτΛ࣮ߦ͢Δࡍʹ Լਤͷ֬ೝ͕ඞཁͰ͢ ɻ PyML in
Niigata SECURITY WARNING 11
PyML in Niigata TEXT CLASSIFICATION 12
PyML in Niigata WORD EMBEDDING 13 <START> this film was
just brilliant casting … with us all [1, 14, 22, 16, 43, 530, 973, …, 19, 178, 32] a1 b1 ⋮ p1 a2 b2 ⋮ p2 a3 b3 ⋮ p3 am bm ⋮ pm am bm ⋮ cm Embedding GlobalAveragePooling1D
PyML in Niigata TIPS HOW TO BRING YOUR OWN DATA
https://colab.research.google.com/notebooks/io.ipynb 14 ԼهͷίʔυΛ࣮ߦ͢ΔͱೝূϦϯΫͱೖྗϑΟʔϧυ͕දࣔ͞Ε·͢ɻ ೝূϦϯΫΛΫϦοΫͯ͠ɺGoogle Colab͔ΒͷGoogle DriveͷΞΫη εΛڐՄ͠ɺൃߦ͞ΕͨτʔΫϯΛೖྗϑΟʔϧυʹϖʔετ͠Enter ΩʔΛԡ͢ͱɺGoogle Drive ͕Ϛϯτ͞Ε·͢ɻ
PyML in Niigata NEXT STEP ଞͷϊʔτϒοΫͬͯΈΔ ϊʔτϒοΫΛίϐʔͯ͠ϞσϧΛ͍ͬͯ͡ΈΔ ॻ੶WEB্ͷίʔυΛ࣮ߦͯ͠ΈΔ 15