Intel AI Analytics Toolkit mo Linux

Fa'amatalaga o oloa

O le AI Kit o se meafaigaluega e aofia ai le tele o siosiomaga conda mo aʻoaʻoga masini ma galuega loloto aʻoaʻoga. E aofia ai siosiomaga mo TensorFlow, PyTorch, ma Intel oneCCL Bindings. E mafai ai e tagata faʻaoga ona faʻapipiʻi a latou faiga e ala i le faʻatulagaina o fesuiaiga o le siosiomaga, faʻaaoga Conda e faʻaopoopo ai afifi, faʻapipiʻi taʻavale ata, ma faʻagata hangcheck. O le meafaigaluega e mafai ona faʻaogaina i le Command Line Interface (CLI) ma e mafai ona faigofie ona tuʻufaʻatasia i galuega o loʻo iai e aunoa ma ni suiga faʻapitoa.

Fa'aogaina o Mea

1. Fa'atulaga lau faiga e ala i le fa'atulagaina o suiga ole si'osi'omaga a'o le'i fa'aauau.
2. Ina ia galue i se Poloaiga Line Interface (CLI), faʻaoga le setvars.sh script e faʻapipiʻi ai meafaigaluega i le oneAPI toolkits e ala i fesuiaiga o le siosiomaga. E mafai ona e maua mai le setvars.sh script e tasi i le vasega po'o taimi uma e te tatalaina ai se faamalama fou. E mafai ona maua le script setvars.sh i le pusa a'a o lau fa'apipi'i oneAPI.
3. Fa'agaoioia siosiomaga conda eseese pe a mana'omia e ala i le fa'atonuga "conda activate ”. O le AI Kit e aofia ai si'osi'omaga conda mo TensorFlow (CPU), TensorFlow ma Intel Extension mo Sample TensorFlow (GPU), PyTorch ma le Intel Extension mo PyTorch (XPU), ma Intel oneCCL Bindings mo PyTorch (CPU).
4. Su'esu'e i si'osi'omaga ta'itasi e feso'otai i le Amataina Sample feso'ota'i i le laulau o lo'o tu'uina atu i totonu o le tusi lesona mo nisi fa'amatalaga i le fa'aogaina o si'osi'omaga ta'itasi.

O fa'atonuga nei e fa'apea ua e fa'apipi'iina le polokalama Intel® oneAPI. Fa'amolemole va'ai le itulau Intel AI Analytics Toolkit mo filifiliga fa'apipi'i. Mulimuli i laasaga nei e fausia ma tamoʻe peiampma le Intel® AI Analytics Toolkit (AI Kit):

1. Fa'atonu lau faiga.
2. Fausia ma faʻatautaia se Sample.

FAAMANATU: O faʻapipiʻi masani Python e fetaui lelei ma le AI Kit, ae o le Intel® Distribution for Python* e sili.
Leai se suiga fa'apitoa i au poloketi o lo'o iai e mana'omia e amata fa'aoga ai i lenei pusa meafaigaluega.

Vaega o Lenei Meafaigaluega

O le AI Kit e aofia ai

• Intel® Optimization mo PyTorch*: O le Intel® oneAPI Deep Neural Network Library (oneDNN) o loʻo aofia i le PyTorch o le faʻaaogaina o le math kernel library mo aʻoaʻoga loloto.
• Intel® Extension mo PyTorch: Intel® Extension mo PyTorch* fa'alautele atu le PyTorch* agava'a fa'atasi ai ma foliga fa'aonaponei ma fa'ata'ita'iga mo se fa'aopoopoga fa'aopoopo i mea faigaluega a le Intel.
• Intel® Optimization mo TensorFlow*: O lenei fa'aliliuga e tu'ufa'atasia ai mea muamua mai le oneDNN i totonu o le TensorFlow taimi mo le fa'avavevaveina o fa'atinoga.
• Intel® Extension mo TensorFlow: Intel® Extension mo TensorFlow* o se mea fa'aopoopo, maualuga le fa'atinoga o le fa'alauteleina o a'oa'oga loloto fa'atatau ile TensorFlow PluggableDevice interface. O lenei faʻaopoopoga faʻaopoopoga e aumaia ai masini Intel XPU (GPU, CPU, ma isi) i totonu o le TensorFlow open source community mo le faʻavaveina o galuega a AI.
• Intel® Distribution for Python*: Maua vave le fa'atinoga o talosaga Python mai le atigipusa, fa'atasi ai ma ni suiga la'ititi pe leai fo'i i lau code. O lenei tufatufaga ua tu'ufa'atasia ma Intel® Performance Libraries e pei o le Intel® oneAPI Math Kernel Library ma le Intel®oneAPI Data Analytics Library.
• Intel® Distribution of Modin* (na'o Anaconda e maua), lea e mafai ai ona e fuaina le muai fa'agasolo i luga o le tele o pona e fa'aaoga ai lenei faletusi atamai, tufatufaina fa'amaumauga fa'amaumauga ma se API tutusa i pandas. O lenei tufatufaga e na'o le fa'apipi'iina o le Intel® AI Analytics Toolkit ma le Conda* Package Manager.
• Intel® Neural Compressor : fa'avave ona fa'aogaina fofo fa'ai'uga maualalo i luga o ta'iala o a'oa'oga loloto ta'uta'ua e pei o TensorFlow*, PyTorch*, MXNet*, ma le ONNX* (Open Neural Network Exchange) ta'i taimi.
• Intel® Extension mo Scikit-a'oa'o*: Ose auala e fa'avavevave ai lau talosaga Scikit-a'oa'o e fa'aaoga ai le Intel® oneAPI Data Analytics Library (oneDAL).
  Patching scikit-a'oa'o ua avea ma auivi a'oa'oga masini e fetaui lelei mo le taulimaina o faafitauli moni o le olaga.
• XGBoost Optimized by Intel: O lenei ta'uta'ua masini-a'oa'oga afifi mo gradient-boosted fa'ai'uga la'au e aofia ai seamless, pa'u-i faatelevaveina mo Intel® tusiata fale e matua faatelevaveina aʻoaʻoga faʻataʻitaʻiga ma faʻaleleia le saʻo mo faʻamatalaga sili atu.

Fa'atulaga lau Faiga - Intel® AI Analytics Toolkit

Afai e te le'i fa'apipi'iina le AI Analytics Toolkit, tagai ile Fa'apipi'i ole Intel® AI Analytics Toolkit. Ina ia fetuutuunai lau faiga, seti suiga ole siosiomaga a'o le'i fa'aauau.

 

Seti Su'esu'ega Si'osi'omaga mo Atina'e CLI
Mo le galue i le Command Line Interface (CLI), o mea faigaluega i le oneAPI toolkits e faʻapipiʻiina e ala i
fesuiaiga o le siosiomaga. Le setiina o fesuiaiga o le siosiomaga e ala i le suʻeina o le seti seti:

Filifiliga 1: Source setvars.sh tasi i le vasega
Source setvars.sh i taimi uma e te tatalaina ai se faamalama faʻailoga fou:

E mafai ona e mauaina le setvars.sh script i le root folder o lau OneAPI installation, lea e masani lava /opt/intel/oneapi/ mo faʻapipiʻi lautele ma ~/intel/oneapi/ mo faʻapipiʻi tumaoti.

Mo faʻapipiʻi lautele (manaʻomia aʻa poʻo sudo avanoa):

• . /opt/intel/oneapi/setvars.sh

Mo mea fa'apipi'i tumaoti:

• . ~/intel/oneapi/setvars.sh

Filifiliga 2: Tasi taimi seti mo setvars.sh
Ina ia otometi ona seti le siosiomaga mo au galuega faatino, ia aofia ai le puna o le poloaiga
/setvars.sh i se faʻamatalaga amata lea o le a faʻaogaina otometi (sui
fa'atasi ai ma le ala i lau oneAPI fa'apipi'i nofoaga). O nofoaga fa'apipi'i fa'aletonu o /opt/
intel / oneapi / mo faʻapipiʻi lautele (manaʻomia aʻa poʻo sudo faʻapitoa) ma ~ / intel / oneapi / mo faʻapipiʻi tumaoti.
Mo example, e mafai ona e faaopoopo i ai le puna /setvars.sh fa'atonuga i lau ~/.bashrc po'o ~/.bashrc_profile po'o ~/.profile file. Ina ia fa'amautu tulaga mo fa'amatalaga uma i luga o lau masini, fai se fa'amaumauga e tasi-laina .sh i lau polokalama /etc/profile.d faila e maua mai ai setvars.sh (mo nisi faʻamatalaga, vaʻai faʻamaumauga Ubuntu i luga ole Siosiomaga Variables).

FAAMANATU
O le setvars.sh script e mafai ona pulea e faʻaaoga ai se faʻatulagaga file, lea e aoga tele pe afai e te manaʻomia le amataina o faʻamatalaga patino o faletusi poʻo le tuʻufaʻatasia, nai lo le faʻaogaina o le "fou" version. Mo nisi faʻamatalaga, vaʻai i le Faʻaaogaina o se Faʻatonuga File e Pulea Setvars.sh.. Afai e te manaʻomia le setiina o le siosiomaga i totonu o se atigi e le o POSIX, vaʻai OneAPI Development Environment Setup mo nisi filifiliga faʻaopoopo.

Laasaga Sosoo

• Afai e te le o faʻaaogaina Conda, poʻo le atinaʻeina mo le GPU, Fau ma Faʻataʻitaʻi se Sample Poloketi.
• Mo tagata fa'aoga Conda, fa'aauau i le isi vaega.
• Mo le atinaʻeina o se GPU, faʻaauau i le GPU Tagata faʻaoga

Conda Siosiomaga i lenei Meafaigaluega
E tele siosiomaga conda o loʻo aofia i le AI Kit. O si'osi'omaga ta'itasi o lo'o fa'amatala i le siata o lo'o i lalo. A maeʻa ona e setiina suiga o le siosiomaga i le CLI environment e pei ona faʻatonuina muamua, ona mafai lea ona e faʻagaoioia ni siosiomaga eseese pe a manaʻomia e ala i le poloaiga lenei:

• fa'agaoioia conda

Mo nisi fa'amatalaga, fa'amolemole su'esu'e i si'osi'omaga ta'itasi e feso'otai i le Amataina Sample sootaga i le laulau i lalo.

Fa'aoga le Conda Clone Function e fa'aopoopo ai afifi ose tagata e le o a'a
O le Intel AI Analytics toolkit o loʻo faʻapipiʻiina i le oneapi folder, lea e manaʻomia ai aʻa avanoa e pulea ai. Atonu e te manaʻo e faʻaopoopo ma tausia ni afifi fou e faʻaaoga ai le Conda *, ae e le mafai ona e faia e aunoa ma le aʻa. Pe, e mafai ona e maua aʻa ae e te le manaʻo e ulufale i le aʻa upu faʻaulu i taimi uma e te faʻagaoioia ai Conda.

Ina ia pulea lou siosiomaga e aunoa ma le faʻaogaina o aʻa, faʻaoga le Conda clone functionality e faʻapipiʻi ai afifi e te manaʻomia i se pusa i fafo atu o le /opt/intel/oneapi/ folder:

1. Mai le faʻamalama faʻamalama tutusa lea na e tamoe ai setvars.sh, faʻailoa le siosiomaga Conda i luga o lau masini:
   • conda env lisi
     E te va'ai i'uga e pei o lenei:
2. Fa'aoga le galuega fa'a-clone e fa'apipi'i ai le si'osi'omaga i se faila fou. I le exampI lalo ifo, o le siosiomaga fou e igoa ia usr_intelpython ma o le siosiomaga o loʻo faʻapipiʻiina e faʻaigoaina faavae (e pei ona faʻaalia i le ata i luga).
   • conda fatu –igoa usr_intelpython –clone base
     O fa'amatalaga clone o le a aliali mai:

1. Fa'agaoioia le si'osi'omaga fou ina ia mafai ai ona fa'aopoopo afifi. conda fa'agaoioia usr_intelpython
2. Fa'amaonia le si'osi'omaga fou o lo'o ola. conda env lisi
   E mafai nei ona e atiaʻe le faʻaogaina o le Conda environment mo Intel Distribution mo Python.
3. Ina ia fa'agaoioia le TensorFlow* po'o le PyTorch* siosiomaga:

TensorFlow

• conda fa'agaoioia tensorflow

PyTorch

• conda activate pytorch

Laasaga Sosoo

• Afai e te le o atinaʻe mo le GPU, Fausia ma Faʻataʻitaʻi se Sample Poloketi.
• Mo le atinaʻeina o se GPU, faʻaauau i le GPU Tagata faʻaoga.

Tagata fa'aoga GPU
Mo i latou o loʻo atinaʻe i luga o se GPU, mulimuli i laasaga nei:

Fa'apipi'i aveta'avale GPU
Afai na e mulimuli i faʻatonuga i totonu o le Taiala Faʻapipiʻi e faʻapipiʻi ai le GPU Avetaavale, e mafai ona e faaseʻeina lenei laasaga. Afai e te le'i fa'apipi'iina aveta'avale, mulimuli i fa'atonuga o lo'o i totonu o le Ta'iala Fa'apipi'i.

Fa'aopoopo le Tagata Fa'aoga ile Vaega Vitio
Mo le fa'atulagaina o galuega a le GPU, e le masani ona maua e tagata e leai ni a'a (masani) i le masini GPU. Ia mautinoa e faʻaopoopo lau tagata masani masani i le vaega vitio; a leai, o binaries tuufaatasia mo le masini GPU o le a toilalo pe a faia e se tagata masani. Ina ia foia lenei faafitauli, faʻaopoopo le tagata e le o aʻa i le vaega vitio:

• sudo usermod -a -G vitio

Taofi le Hangcheck
Mo talosaga o lo'o i ai le GPU ua leva ona fa'atino galuega fa'atino i totonu o si'osi'omaga moni, fa'agata hangcheck. E le fautuaina lenei mea mo virtualizations poʻo isi faʻaoga masani o le GPU, pei o taaloga.

O se galuega mamafa e sili atu ma le fa sekone mo masini GPU e faʻatino o se galuega umi e alu. E le mafai, o filo ta'ito'atasi e agava'a e avea ma uta umi o galuega e manatu o lo'o tautau ma fa'amutaina. E ala i le fa'agata o le taimi fa'agata hangcheck, e mafai ona e 'alofia lenei fa'afitauli.

FAAMANATU: Afai e fa'afou le fatu, o le hangcheck e otometi lava ona mafai. Fa'ata'ita'i le fa'agasologa o lo'o i lalo pe a uma fa'afouga fatu uma e fa'amautinoa ua fa'aletonu le hangcheck.

1. Tatala se nofoaga.
2. Tatala le grub file i totonu /etc/default.
3. I le grub file, su'e le laina GRUB_CMDLINE_LINUX_DEFAULT=”” .
4. Ulufale le tusitusiga lenei i le va o upusii (“”):
5. Fa'atonu le poloaiga:
   sudo faʻafouina-grub
6. Toe fa'afou le faiga. Hangcheck e tumau pea le le atoatoa.

Laasaga Sosoo
I le taimi nei ua uma ona e faʻatulagaina lau polokalama, faʻagasolo i le Fausia ma Faʻataʻitaʻi se Sample Poloketi.

Fausia ma faʻatautaia se Sample Fa'aaogaina o le Laina Poloaiga

Intel® AI Analytics Toolkit
I lenei vaega, o le ae faʻatautaia se poloketi faigofie "Hello World" e faʻamasani oe i le faagasologa o le fausiaina o galuega, ona fausia ai lea o lau lava galuega.

FAAMANATU: Afai e te leʻi faʻatulagaina lau siosiomaga atinaʻe, alu i le Faʻatonu lau polokalama ona toe foʻi lea i le itulau lea. Afai ua uma ona e faʻamaeʻaina laasaga e faʻapipiʻi ai lau polokalama, faʻaauau i laasaga o loʻo i lalo.

E mafai ona e faʻaogaina se faʻamalama faʻamalama poʻo le Visual Studio Code * pe a galue mai le laina faʻatonu. Mo fa'amatalaga i le fa'aogaina o le VS Code i le lotoifale, va'ai le Fa'aoga Fa'avae ole Visual Studio Code ma le tasiAPI ile Linux*. Mo le fa'aoga mamao o le VS Code, va'ai le Atina'e Fa'asinoala a le Ata Va'ai mamao ma le tasiAPI ile Linux*.

Fausia ma faʻatautaia se Sample Poloketi
O le sample mea o lo'o i lalo e tatau ona fa'apipi'i i lau faiga a'o le'i mafai ona e fausia le sample poloketi:

Ina ia va'ai i se lisi o vaega e lagolagoina CMake, va'ai Fa'aaoga CMake i Talosaga tasiAPI.

Fausia Lau Lava Poloketi
Leai se suiga fa'apitoa i au poloketi Python o lo'o i ai nei e mana'omia e amata fa'aoga ai i lenei mea faigaluega. Mo poloketi fou, o le fa'agasologa e mulimulita'ia le fa'agasologa o lo'o fa'aaogaina mo le fa'atupuina o sample Talofa lalolagi poloketi. Va'ai ile Hello World README files mo faatonuga.

Fa'ateleina Fa'atinoga
E mafai ona e mauaina faʻamaumauga e fesoasoani ia te oe e faʻateleina le faʻatinoga mo TensorFlow poʻo PyTorch.

Fa'atulaga Lou Si'osi'omaga

FAAMANATU: Afai e le o avanoa lau siosiomaga faʻapitoa, pe afai e te manaʻo e faʻaopoopo afifi i lau siʻosiʻomaga mataʻutia, ia mautinoa ua e faʻamaeʻaina laasaga i le Faʻaaogaina o le Conda Clone Function e faʻaopoopo ai afifi e pei o se tagata e le o aʻa.

Afai o lo'o e atia'e i fafo atu o se atigipusa, fa'apipi'i le fa'amatalaga nei e fa'aoga ai le Intel® Distribution for Python*:

• • /setvars.sh
• o fea o le mea lea na e fa'apipi'i ai le pusa meafaigaluega. O le fa'aletonu o le fa'apipi'i fa'atonu o le:
• Root poʻo sudo faʻapipiʻi: /opt/intel/oneapi
• Fa'atonuga fa'aoga fa'apitonu'u: ~/intel/oneapi

FAAMANATU: O le setvars.sh script e mafai ona pulea e faʻaaoga ai se faʻatulagaga file, lea e aoga tele pe afai e te manaʻomia le amataina o faʻamatalaga patino o faletusi poʻo le tuʻufaʻatasia, nai lo le faʻaogaina o le "fou" version. Mo nisi faʻamatalaga, vaʻai i le Faʻaaogaina o se Faʻatonuga File e Pulea Setvars.sh. Afai e te manaʻomia le setiina o le siosiomaga i totonu o se atigi e le o POSIX, vaʻai i le OneAPI Development Environment Setup mo nisi faʻatulagaga filifiliga.

Ina ia fesuia'i si'osi'omaga, e tatau ona e fa'ate'aina muamua le si'osi'omaga gaioi.
O le example faʻaalia le faʻatulagaina o le siosiomaga, faʻagaoioia TensorFlow *, ona toe foʻi lea i le Intel Distribution mo Python:

La'u mai se Container

Intel® AI Analytics Toolkit
O koneteina e fa'atagaina oe e fa'atutu ma fa'atulaga si'osi'omaga mo le fau, fa'agaioia ma fa'ata'ita'iina o talosaga a le oneAPI ma fa'asoa fa'aaoga ata:

• E mafai ona e faʻapipiʻi se ata o loʻo i ai se siosiomaga na faʻapipiʻiina muamua ma meafaigaluega uma e te manaʻomia, ona atiaʻe lea i totonu o lena siosiomaga.
• E mafai ona e faʻasaoina se siosiomaga ma faʻaoga le ata e faʻanofo ai lena siosiomaga i se isi masini e aunoa ma se seti faʻaopoopo.
• E mafai ona e saunia ni koneteina ma seti eseese o gagana ma taʻaloga, meafaigaluega auʻiliʻili, poʻo isi meafaigaluega, pe a manaʻomia.

La'u Ata Docker*
E mafai ona e siiina se ata Docker * mai le Containers Repository.

FAAMANATU: O le ata Docker e ~ 5 GB ma e mafai ona ave ~ 15 minute e sii mai ai. O le a manaʻomia le 25 GB o le avanoa disk.

1. Fa'amatala le ata:
   ata=intel/oneapi-aikit docker toso “$image”
2. Toso le ata.
   toso e faife'au "$ata"

O le taimi lava e sii mai ai lau ata, faʻagasolo i le Faʻaaogaina o Containers ma le Line Line.

Fa'aaogaina o Containers ma le Poloaiga Line
Intel® AI Analytics Toolkit La'u sa'o mai pusa ua uma ona fau. O le fa'atonuga o lo'o i lalo mo le PPU o le a tu'uina atu oe i se fa'atonuga vave, i totonu o le koneteina, i le faiga fa'afeso'ota'i.

PPU
image=intel/oneapi-aikit docker run -it “$image”

Fa'aaogā le Intel® Faufautua, Intel® Inspector po'o le VTune™ fa'atasi ma Containers
A fa'aogaina nei mea faigaluega, e tatau ona tu'uina atu i le koneteina ni mea fa'aopoopo: –cap-add=SYS_ADMIN –cap-add=SYS_PTRACE

• tamo'e faifa'ato'aga –cap-add=SYS_ADMIN –cap-add=SYS_PTRACE \ –device=/dev/dri -it “$image”

Fa'aaogaina Cloud CI Systems

Cloud CI e fa'ataga ai oe e fau ma fa'ata'ita'i otometi lau polokalama. Vaʻai le repo ile github mo example fa'atulagaina fileo lo'o fa'aogaina le tasiAPI mo faiga fa'apitoa o le ao CI.

Fa'afitauli mo le Intel® AI Analytics Toolkit

Fa'asilasilaga ma Fa'asalaga

E ono mana'omia e tekinolosi a le Intel ni meafaigaluega fa'aagaoioi, polokalama po'o le fa'agaoioia o auaunaga. E leai se oloa po'o se vaega e mafai ona matua malupuipuia.
E ono fesuisuiai au tau ma taunuuga.

© Intel Corporation. Intel, le Intel logo, ma isi fa'ailoga Intel o fa'ailoga fa'ailoga a le Intel Corporation po'o ona lala. O isi igoa ma fa'ailoga e mafai ona ta'ua o se meatotino a isi.

Oloa ma Fa'amatalaga Fa'atinoga

Fa'atinoga e eseese ile fa'aoga, fa'atulagaina ma isi mea. A'oa'o atili ile www.Intel.com/PerformanceIndex.
Fa'aaliga toe teuteu #20201201

Leai se laisene (fa'aalia pe fa'aali, e le estoppel po'o se isi mea) i so'o se aia tatau tau le atamai ua fa'atagaina e lenei pepa. O oloa o loʻo faʻamatalaina e ono iai ni faʻaletonu mamanu poʻo ni mea sese e taʻua o le errata e ono mafua ai ona alu ese le oloa mai faʻamatalaga faʻasalalau. O fa'ailoga fa'apitoa o lo'o iai nei e avanoa pe a talosaga.

E fa'afitia uma e Intel fa'apolopolo fa'amaonia, e aofia ai e aunoa ma se fa'atapula'aina, fa'amaonia fa'amaonia o le fa'atau, talafeagai mo se fa'amoemoe fa'apitoa, ma le le solia, fa'apea fo'i ma so'o se saisai e tula'i mai i le fa'atinoga, faiga tau fefa'ataua'iga, po'o le fa'aogaina i fefa'ataua'iga.

Pepa / Punaoa

Intel AI Analytics Toolkit mo Linux [pdf] Taiala mo Tagata Fa'aoga AI Analytics Toolkit mo Linux, AI Analytics Toolkit, Analytics Toolkit mo Linux, Analytics Toolkit, Toolkit

Fa'asinomaga

• Tusi Taiala