Upgrade to Pro — share decks privately, control downloads, hide ads and more …

mlflowをwindowsで起動するのはめんどくさい

F0e950c6bf20815f3d169f063d3c5d4f?s=47 Aipa
March 15, 2019
Technology
0
660

 mlflowをwindowsで起動するのはめんどくさい

データ分析チャンプルー val.1@琉球大学で発表してきた

https://nobdata.connpass.com/event/122314/

F0e950c6bf20815f3d169f063d3c5d4f?s=128

Aipa

March 15, 2019
Tweet

More Decks by Aipa

See All by Aipa
pyconjp 2021 振り返り
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
29
ジュラシック・アイパー 2022 怒りのデスフライト
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
120
（せーのっ）「いますぐはじめるえむえるおぷす」
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
130
Gitlab-RunnerからDockerコンテナを起動して テストする構成を用意した
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
88
#club_家庭菜園_進捗報告LT
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
40
サーバーレスピタゴラスイッチ
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
22
BigQueryで金を溶かさない知識
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
370
ゾンビでわかる分類評価指標
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
1
110
LT作成を手伝った話（愚痴）
F0e950c6bf20815f3d169f063d3c5d4f?s=48 commander_aipa
0
600

Other Decks in Technology

See All in Technology
Microsoft Data Analytics trends : ”Lakehouse” , ”Data Mesh"
0d7a3b61c5c1f64a50136cf4bb5702b7?s=48 ryomaru0825
2
100
eBPF-based Container Networking
98b5759510e487ade4145247bc856366?s=48 johnlin
2
1.1k
psql, my favorite tool!
18749909c147a9ee98f6b112911943cc?s=48 nuko_yokohama
1
170
Settlement simulation testing to ensure correct settlement processing
23e9a5be8e46efd1cf6a07694a047237?s=48 applepine1125
2
680
一番小さなリファクタリングの話
Bf7fe621f4fe1615c228ef8a79b87282?s=48 shirayanagiryuji
0
140
DevelopersIO 2022 俺のTerraform Pipeline
08dd0b93e011904f40d60d5a1b54c671?s=48 takakuni
0
420
EC／CRMの自社サービス開発をマネジメントするようになって1年でやってきたこととこれから / devio2022-takano-sho-road-to-good-development-team-management
Fee058832252e72722b8b03073ff6324?s=48 masaru_b_cl
0
360
金融スタートアップの上場準備で大事にしたマインドセット / 2022-08-04-the-mindset-in-preparing-for-ipo
0251532f4fb413ea1a026efe6eb16b87?s=48 stajima
0
290
AWSで実現する「好き」の感情 / Develop Suki by AWS #devio2022
1e72c797afb9bab65655a1eb7d183460?s=48 syobochim
1
230
もうつまらないとは言わせない！「わかりやすい」プレゼンを作るために気をつけたいこと
E97091ac0d2c69ce9f9e8a8d0a2ea066?s=48 torisoup
4
1.3k
JAWS-UG 朝会 #36 登壇資料
08dd0b93e011904f40d60d5a1b54c671?s=48 takakuni
1
500
Goで実装するブランドネットワークとの接続ポイント
E10d51de51c08e913216385be87dc1ca?s=48 pongzu
2
240

Featured

See All Featured
Become a Pro
828ace851b606e1206900f26459f55ad?s=48 speakerdeck
PRO
3
900
Rebuilding a faster, lazier Slack
C0b42577cfc2b321be3474618107e933?s=48 samanthasiow
62
7.3k
Scaling GitHub
78b475797a14c84799063c7cd073962f?s=48 holman
451
140k
Designing for Performance
245cee81a9c424266e5e401d844ea881?s=48 lara
597
63k
Pencils Down: Stop Designing & Start Developing
79acae287842acf29084005533a7fb3b?s=48 hursman
113
9.8k
How to train your dragon (web standard)
053e75a5b48b44d6dd0612795dfb326d?s=48 notwaldorf
60
3.9k
Web Components: a chance to create the future
E190023b66e2b8aa73a842b106920c93?s=48 zenorocha
303
40k
Support Driven Design
1519587b1a0febb658c36c94f6272b0d?s=48 roundedbygravity
87
8.6k
KATA
E9221b172e78e888355fa2fe4541b595?s=48 mclloyd
7
8.8k
The Brand Is Dead. Long Live the Brand.
A415db3ac9e9668acbc1fb36eead84ac?s=48 mthomps
46
2.7k
WebSockets: Embracing the real-time Web
E76911cbe088e5b850d966de3fc7435b?s=48 robhawkes
57
5.5k
The Invisible Side of Design
B3d6434763caa0ef5dc4b792662c49f7?s=48 smashingmag
290
48k

Transcript

  1. σʔλ෼ੳͪΌΜ΀Δʔ vol.1 ΞΠύʔୂ௕@ླྀٿେֶ

  2. ࣗݾ঺հ

  3. ࣗݾ঺հ • ΞΠύʔୂ௕(ࠓ೥ 29) • ೋ೔ਲ͍Ͱ͢ • ͪΎΒσʔλגࣜձࣾ • PyData.OkinawaڞಉΦʔΨφΠβʔ

    • ࡢ೔͔ΒϏʔϧ͕ݏ͍ʹͳΓ·ͨ͠

  4. ࠂ஌

  5. PyData.Okinawa #38 LTେձ https://pydataokinawa.connpass.com/event/121724/ ໌೔ 15:00 ~ 18:00 ৔ॴ͸ླྀٿେֶʢ஍Ҭ૑ੜ૯߹ݚڀ౩ʣ

  6. ঎඼͕͋Γ·͢ʂʂʢશһ౤ථʣ

  7. ݱ࣌఺શһೖ৆֬ఆ ʢΑ͔ͬͨΒཔΉདྷͯ͘ΕཔΉʣ

  8. ΅͘΋͜Ε͔Β࡞Γ·͢ʂʂ ʢ͔ͩΒ҆৺ͯ͠དྷͯ͘Εʣ

  9. ࿩͢લʹ

  10. ຊεϥΠυͰ͸ಡΈͮΒ͍Լهϫʔυ͕Ͱ͖ͯ·͢ • fcntl • WSGI • Gunicorn

  11. ຊεϥΠυͰ͸͜͏ݺͼ·͢ • fcntl → ͑;͜ΜͱΖʔΔ • WSGI → ͏͌ͣ͗ʔ •

    Gunicorn → ͙ʹ͜ʔΜ

  12. ຊεϥΠυͰ͸͜͏ݺͼ·͢ • fcntl → ͑;͜ΜͱΖʔΔ • WSGI → ͏͑͗͢ •

    Gunicorn → ͙ʹ͜ʔΜ https://twitter.com/moon_in_nagano/status/1001315017273167872

  13. ຊεϥΠυͰ͸͜͏ݺͼ·͢ • fcntl → ͑;͜ΜͱΖʔΔ • WSGI → ͏͑͗͢ •

    Gunicorn → ͙ʹ͜ʔΜ https://twitter.com/moon_in_nagano/status/1001315017273167872

  14. ຊ೔͓࿩͢Δ͜ͱ

  15. ຊ೔͓࿩͢Δ͜ͱ

  16. mlflowྑͦ͞͏ https://qiita.com/ike_dai/items/3121a8cc8398c6ec6a33

  17. ͔͠͠ࠓճͷཁ݅ • Windows OS • HogeHogeFlow • FugaFuga-learn

  18. ͔͠͠ࠓճͷཁ݅ • Windows OS • HogeHogeFlow • FugaFuga-learn

  19. do not support (´ʀωʀʆ)

  20. ֬ೝͯ͠ͳ͍͚ͩͰಈ͘Ͱ͠ΐʢ͸ͳ΄͡ʔ

  21. (´ʀωʀʆ) GDOUM͕ͳ͍

  22. WindowsͰ͸fcntl͸ಈ͔ͳ͍ https://stackoverflow.com/questions/1422368/fcntl-substitute-on-windows

  23. WindowsͰ͸fcntl͸ಈ͔ͳ͍ https://stackoverflow.com/questions/1422368/fcntl-substitute-on-windows

  24. Կނಈ͔ͳ͍͔ௐ΂Δ

  25. Τϥʔ͔Βௐ΂Δ

  26. Τϥʔ͔Βௐ΂Δ https://github.com/benoitc/gunicorn/blob/master/gunicorn/util.py#L8

  27. gunicornͰΤϥʔ͕ग़͍ͯΔΑ͏ͩ • gunicorn is Կʁ • Python༻ͷWSGI HTTP Server

  28. gunicornΛwindowsͰ ಈ࡞͢ΔΑ͏ʹ मਖ਼͢Ε͹͍͍ͷͰ͸??!

  29. Dead end. ×2

  30. ͓ʁʢ^ω^ = ^ω^ʣ

  31. Dead end. ×3

  32. Dead end. ×3

  33. mlflow͸ windowsͰ͸࢖͑·ͤΜ

  34. ɹ΋͏গ͠ؤுͬͯΈΔ

  35. mlflowͷίʔυΛಡΉ HVOJDPSOΛ ىಈ͢Δͱ͜Ζ NMqPXTFSWFS Λ࣮ߦ͢Δͱݺ͹ΕΔؔ਺ https://github.com/mlflow/mlflow/blob/master/mlflow/server/__init__.py

  36. mlflowͷίʔυΛಡΉ https://github.com/mlflow/mlflow/blob/master/mlflow/server/__init__.py 'MBTL࢖ͬͯΔ

  37. ݟ͖͑ͯͨΞʔΩςΫνϟʢΠϯϑϥʣ

  38. ݟ͖͑ͯͨΞʔΩςΫνϟʢαʔόʣ

  39. WSGI͍Βͳ͍આ ᶃHVOJDPSOͷྶѹΛফ͢ ᶄ'MBTLΛىಈ͢Δ͚ͩ

  40. վम͢Δ

  41. mlflow/server/__init__.pyΛ Flask͕ىಈ͢ΔΑ͏मਖ਼ ͓·͡ͳ͍ ˞ىಈޙʹ؀ڥม਺͕ ͳ͍ͱΤϥʔʹͳΔ 'MBTLͷىಈίʔυ

  42. ىಈͨ͠

  43. ىಈͨ͠

  44. ·͊Ͱ΋Flask͚ͩ͡Ό ෺଍Γͳ͍৔߹͕͋Δ

  45. FlaskͷσϑΥϧτͰ͸ಉ࣌ΞΫηεΛॲཧͰ͖ͳ͍ https://qiita.com/5zm/items/251be97d2800bf67b1c6

  46. ͳͥgunicornͳͲͷWSGIαʔόʔΛ࢖͏ͷ͔ https://blog.hirokiky.org/entry/2018/10/06/151830

  47. ΋͠ຊ൪Քಇ͍ͤͨ͞ͳΒ Apache, Nginx౳Λ࢖͍͍ͨ

  48. WindowsͰಈ࡞͢Δ WSGIΛ୳͢

  49. ͋ͬͨ https://docs.pylonsproject.org/projects/waitress/en/stable/

  50. ΊͬͪΌ؆୯ͦ͏ https://docs.pylonsproject.org/projects/waitress/en/stable/usage.html

  51. WSGIΛࠩ͠ସ͑Δઆ ᶃHVOJDPSOΛ XBJUSFTT΁มߋ͢Δ waitress

  52. मਖ਼͢Δ

  53. mlflow/server/__init__.py 'MBTLͷΦϒδΣΫτΛ XBJUSFTTʹ৯ΘͤΔ͚ͩ

  54. ىಈͨ͠

  55. ࡶײ

  56. ࡶײ • mlflow͸windowsͰ΋ಈ͔ͤΔ • શػೳςετͨ͠Θ͚Ͱ͸ͳ͍ͷͰϋϚΓͲ͜Ζ͸ ·ͩ͋Γͦ͏ • Ͱ΋ɺ΍ͬͺΓΊΜͲ͍͔͘͞Β
 Docker࢖ͬͨ΄͏͕͍͍Α

  57. None