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
AWS GPUインスタンスの使い方と注意点
Search
Sponsored
·
Ship Features Fearlessly
Turn features on and off without deploys. Used by thousands of Ruby developers.
→
tominaga443
July 01, 2016
Technology
4.8k
0
Share
AWS GPUインスタンスの使い方と注意点
2016/06/30 SD Study
tominaga443
July 01, 2016
More Decks by tominaga443
See All by tominaga443
Amazon Machine Learning Concepts
tominaga443
2
460
Other Decks in Technology
See All in Technology
RubyでRuby拡張を書いたらRubyより35倍速になったってどういうこと??
kazuho
1
130
既存プロダクトQAから新規プロダクトQAへ
ryotakahashi
0
170
Claude Code / Codex / Kiro に AWS 権限を 渡すとき、何を設計すべきか
k_adachi_01
6
1.9k
Fラン学生が考える、AI時代のデザインに執着した突破口
husengs7
1
230
Directions Asia 2026 | Beyond Buildable AI Agents: Let’s Visualize Partner Value in the AI Era
ryoheig0405
0
130
TypeScriptはどのようにどこまで推論できるのか ─ とにかく as は禁止で
ypresto
1
280
生成AI時代に信頼性をどう保ち続けるか - Policy as Code の実践
akitok_
1
530
Cortex(Code) を ML モデルの 精度改善サイクルに組み込む.pdf
oimo23
0
250
AI時代に求められる思考のパラダイムシフト
nrinetcom
PRO
0
110
Geek Woman の育ち方 〜コミュニティとAIと〜
chicaco
0
160
20260515 OpenIDファウンデーション・ジャパンご紹介
oidfj
0
260
JaSSTに関わることで変わった人生観 #jasstnano
makky_tyuyan
0
170
Featured
See All Featured
Noah Learner - AI + Me: how we built a GSC Bulk Export data pipeline
techseoconnect
PRO
0
180
Navigating the moral maze — ethical principles for Al-driven product design
skipperchong
2
370
Raft: Consensus for Rubyists
vanstee
141
7.4k
<Decoding/> the Language of Devs - We Love SEO 2024
nikkihalliwell
1
210
New Earth Scene 8
popppiees
3
2.2k
A brief & incomplete history of UX Design for the World Wide Web: 1989–2019
jct
2
370
From Legacy to Launchpad: Building Startup-Ready Communities
dugsong
0
210
Done Done
chrislema
186
16k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
35
2.4k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
37
6.4k
Collaborative Software Design: How to facilitate domain modelling decisions
baasie
1
220
Getting science done with accelerated Python computing platforms
jacobtomlinson
2
200
Transcript
"84(16Πϯελϯε ͍ํͱҙ !UPNJOBHB
ࣗݾհ w ໊લɿӬળࢹʢͱΈͳ͕Α͠Έʣ w ॴଐɿ5*4גࣜձࣾ ɹɹɹઓུٕज़ηϯλʔ"*ٕज़ਪਐࣨ w ܦྺɿ w ઌՊֶٕज़େֶӃେֶଔۀ
w ݄5*4ೖࣾઓུٕज़ηϯλʔଐ ΫϥυΠϯϑϥʹؔ͢Δݚڀ։ൃʹैࣄ w ݄ΑΓಉ"*ٕज़ਪਐࣨ ΤʔδΣϯτରͷݚڀ։ൃΛ։࢝ ஶॻɿ"84ΤΩετύʔτཆಡຊ ػցֶशฤ ʢٕज़ධࣾʣ
ΞδΣϯμ w "84ʹ͍ͭͯ w (16Πϯελϯεͷ࡞ํ๏ w (16ΠϯελϯεͰ$IBJOFSΛಈ࡞ͤ͞Δ w ར༻্ͷҙ
"NB[PO8FC4FSWJDF w "NB[PODPNͷΫϥυΠϯϑϥαʔϏεʢ*BB4ʣ w αʔόΠϯϑϥΛ͖ͳ࣌ʹඞཁͳ͚ͩར༻Ͱ͖Δ w ଟ༷ͳαʔϏεΛఏڙ w ԾϚγϯɺετϨʔδɺσʔλϕʔε ωοτϫʔΫ੍ޚɺϝʔϧ৴ɺػցֶशͳͲ
w Ϧʔδϣϯ͝ͱʹར༻ՄೳͳαʔϏεྉ͕ۚҟͳΔ ࢀߟʲॳ৺ऀ͚ʳ"84ΛֶͿͨΊͷϦϯΫूʢ݄൛ʣIUUQEFWDMBTTNFUIPEKQDMPVEBXTBXTMFBSOJOH
"NB[PO&$ w ΠϯελϯεʢԾϚγϯʣΛఏڙ͢ΔαʔϏε w 04ηοτΞοϓࡁΈͷԾϚγϯΛଈ࣌ʹར༻Մೳ w ར༻ྉۚ࣌ؒ୯ҐͰͷ՝ۚ ʢ࡞͙ͯ͢͠ʹআͯ࣌ؒ͠ͷྉ͕ۚൃੜʣ w ఀࢭதͷΠϯελϯεྉ͕ۚൃੜ͠ͳ͍
w 04ͷϥΠηϯεྉɺσΟεΫ༰ྔͳͲผ్՝ۚ ࢀߟɹ&$ར༻ྉۚɹIUUQTBXTBNB[PODPNKQFDQSJDJOH
(16Πϯελϯε w /7*%*"ͷ(16Λࡌ ʢ$6%"ίΞ (#ϝϞϦʣ(16 w छྨͷΠϯελϯελΠϓʢεϖοΫʣ w HYMBSHF (16
W$16 (#ϝϞϦʢI!౦ژʣ w HYMBSHF (16 W$16 (#ϝϞϦʢI!౦ژʣ
(16Πϯελϯεͷ࡞ w $6%"ΠϯετʔϧࡁΈ".* w Πϯελϯεͷ࡞ํ๏ w ".* w 04ͱιϑτΣΞߏΛ هͨ͠ςϯϓϨʔτ
$6%"ΠϯετʔϧࡁΈ".*ʢʣ w /7*%"ఏڙͷ".*$6%"ΠϯετʔϧࡁΈ "NB[PO-JOVYͱ8JOEPXT4FSWFS͕ར༻Մೳ
$6%"ΠϯετʔϧࡁΈ".*ʢʣ w 6CVOUV#JUGVTJPOͷ".*Ͱར༻Մೳ ιϑτΣΞར༻ྉ͋ΓʢΠϯελϯεྉۚ ʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
Πϯελϯεͷ࡞ʢʣ
$IBJOFSΛಈ࡞ͤ͞Δ w DV%//ͷΠϯετʔϧ w $IBJOFSͷΠϯετʔϧ w αϯϓϧϓϩάϥϜͷ࣮ߦ
DV%//ͷऔಘ w /7*%*"ͷσϕϩούʔαΠτʹొ w IUUQTEFWFMPQFSOWJEJBDPNDVEOO w ొྃޙμϯϩʔυϖʔδ͔ΒDV%//Λऔಘ w ݱߦͷ҆ఆ൛ DV%//W
.BZ GPS$6%" w ࠓճ-JOVY༻ͷϥΠϒϥϦΛऔಘ DV%//W-JCSBSZGPS-JOVY
DV%//ͷΠϯετʔϧ $6%"όʔδϣϯͷ֬ೝʢόʔδϣϯ͕ߋ৽͞Ε͍ͯͨΒରԠ͢ΔύοέʔδΛऔಘ͢Δʣ (16JOTUBODF DBUPQUOWJEJBDVEBWFSTJPOUYU $6%"7FSTJPO μϯϩʔυͨ͠DV%//ύοέʔδͷసૹ -PDBM1$ TDQJLFZQBJSDVEOOMJOVYYWHBUH[FDVTFS!JOTUBODFJQd
(16JOTUBODF MTd DVEOOMJOVYYWHBUH[ DV%//ͷΠϯετʔϧ (16JOTUBODF UBS[YWGDVEOOMJOVYYWHBUH[ TVEPDQDVEBJODMVEFDVEOOIPQUOWJEJBDVEBJODMVEF TVEPDQDVEBMJC PQUOWJEJBDVEBMJC
$IBJOFSͷΠϯετʔϧ $IBJOFSͷΠϯετʔϧ TVEP$6%"@1"5)PQUOWJEJBDVEBQJQJOTUBMMDIBJOFS $6%"αϙʔτͷ֬ೝʢDVQZ͕ΠϯϙʔτͰ͖Δʣ QZUIPO JNQPSUDVQZ )%'ʹରԠ͢Δ߹ TVEPZVNDPOGJHNBOBHFSFOBCMFFQFM TVEPZVNJOTUBMMZIEGEFWFM
TVEPQJQJOTUBMMIQZ
αϯϓϧΛಈ࡞ͤ͞Δ $IBJOFSͷ(JUϦϙδτϦΛऔಘ TVEPZVNJOTUBMMZHJU HJUDMPOFIUUQTHJUIVCDPNQGOFUDIBJOFSHJU αϯϯϓϧͷ./*45Λ࣮ߦ UJNFQZUIPODIBJOFSFYBNQMFTNOJTUUSBJO@NOJTUQZHQV (16 VOJU .JOJCBUDITJ[F FQPDI
/FUXPSLUZQFTJNQMF ʜ FQPDI USBJONFBOMPTT BDDVSBDZ UISPVHIQVUJNBHFTTFD UFTUNFBOMPTT BDDVSBDZ SFBM NT VTFS NT TZT NT
αϯϓϧΛಈ࡞ͤ͞Δ $IBJOFSͷ(JUϦϙδτϦΛऔಘ TVEPZVNJOTUBMMZHJU HJUDMPOFIUUQTHJUIVCDPNQGOFUDIBJOFSHJU αϯϯϓϧͷ./*45Λ࣮ߦ UJNFQZUIPODIBJOFSFYBNQMFTNOJTUUSBJO@NOJTUQZHQV (16 VOJU .JOJCBUDITJ[F FQPDI
/FUXPSLUZQFTJNQMF ʜ FQPDI USBJONFBOMPTT BDDVSBDZ UISPVHIQVUJNBHFTTFD UFTUNFBOMPTT BDDVSBDZ SFBM NT VTFS NT TZT NT ./*45NT
ར༻্ͷҙ w ॏཁͳҙࣄ߲ w '"2
Ϋϥυഁ࢈ʹؾΛ͚ͭ·͠ΐ͏ w ར༻͠ͳ͍ͱ͖ඞͣఀࢭ͢Δ w ԋࢉతͰͳ͍(16ΠϯελεΛ࡞͠ͳ͍ w ֬ೝআΛࣗಈԽ͢Δ ࢀߟ-BNCEBͷ4DIFEVMFΠϕϯτͰ&$Λࣗಈىಈˍࣗಈఀࢭͯ͠Έͨ IUUQEFWDMBTTNFUIPEKQDMPVEBXTMBNCEBTDIFEVMFEFWFOUVQEPXO (16
UNJDSP I Y HYMBSHF I
'"2 w $6%"ΛࣗͰಋೖ͍ͨ͠ w 6CVOUV IUUQUMFZEFOHJUIVCJPCMPHDVEBEPUPOBXTHQVJOTUBODFSVOOJOH VCVOUVEPU w "NB[PO-JOVY IUUQFZFCBMMPOMZDPNCMPHDVEBPQFODMBNB[PO
w 5FOTPSqPXΛ͍͍ͨ w (16Πϯελϯεʹࡌ͞Ε͍ͯΔ(3*%,ಛผͳର Ԡ͕ඞཁ IUUQRJJUBDPNIBJUFNTEF
'"2 w εϙοτΠϯελϯεͩͱ͍҆Β͍͠ w ݱࡏ(16ΠϯελϯεधཁաଟͷͨΊ҆ఆ͠ͳ͍ w ௨ৗͰɺఀࢭͰ͖ͳ͍ͷͰ݁Ռతʹߴͭ͘͘Մೳੑ͋Γ w σϑΥϧτͰ͍ͭͯ͘Δ44%(#Λ͍͍ͨ w
NOUʹׂΓͯΒΕ͍ͯΔ w ఀࢭ࣌ʹอଘ͞Εͳ͍ͨΊσʔλஔ࣌ҙ w Πϯελϯε࡞࣌ͷσΟεΫ༰ྔΛେ͖Ίʹ͓ͯ͘͠
'"2 w Ҏલ࡞ͨ͠Πϯελϯε͕Ұཡʹແ͍ w ݱࡏͷϦʔδϣϯ͕࡞࣌ͱಉ͔֬͡ೝ͠·͠ΐ͏ w ΞΧϯτڞ༗ͷ߹ॴ༗ऀ͕͔Δ໊લΛ͚Δ w ΠϯελϯεʹΞΫηεͰ͖ͳ͍ w
ηΩϡϦςΟάϧʔϓͰ1PSU͕ڐՄ͞Ε͍ͯΔ͔֬ೝ w 44)ͷ߹ɺΩʔϖΞɺϢʔβ໊͕͍͋ͬͯΔ͔֬ೝ VCVOUVVCVOUVϢʔβɺ"NB[PO-JOVYFDVTFS ʢ".*ʹΑͬͯҟͳΔͷͰ֬ೝ͍ͯͩ͘͠͞ʣ
·ͱΊ w "84Ͱ(16ϚγϯΛଈ࣌ར༻Ͱ͖Δ w ϑϨʔϜϫʔΫʹΑͬͯ$6%"ͷ࠶Πϯετʔϧ͕ඞཁ w ͦΕͳΓʹίετ͕͔͔ΔͷͰඞཁͳͱ͖ͷΈར༻͢Δ
-FU`T5SZ(16