Me pēhea te noho hei Kaiwhakawhanake AI. Ngā Whakarāpopototanga.

Kāore koe i konei mō ngā mahi koretake. Me whai ara mārama koe mō te noho hei Kaiwhakawhanake AI me te kore e totohu ki ngā ripa mutunga kore, ki ngā kupu whakapati, ki te kore rānei e taea te tātari. Pai. Mā tēnei aratohu koe e hoatu ki te mahere pūkenga, ki ngā taputapu e tino hira ana, ki ngā kaupapa e karangahia ana, me ngā tikanga e wehewehe ana i te mahi whakapati me te tuku. Me tīmata tāu hanga.

Ngā tuhinga ka pai pea koe ki te pānui i muri i tēnei:

🔗 Me pēhea te tīmata i tētahi kamupene AI
He aratohu taahiraa-i-te-taahiraa mō te hanga, te putea, me te whakarewa i tō pakihi hou AI.

🔗 Me pēhea te hanga i tētahi AI i runga i tō rorohiko
Akohia te waihanga, te whakangungu, me te whakahaere i ngā tauira AI ā-rohe me te ngāwari.

🔗 Me pēhea te hanga i tētahi tauira AI
He whakamārama whānui mō te hanganga tauira AI mai i te ariā ki te whakatinanatanga.

🔗 He aha te AI tohu
Tūhurahia te mahi a te AI tohu me te hiranga tonu o tēnei mahi i ēnei rā.

He aha te mea e pai ai te Kaiwhakawhanake AI✅

Ehara te kaiwhakawhanake AI pai i te tangata e maumahara ana ki ngā arotau katoa. Engari ko te tangata ka taea te tango i tētahi raruraru pōuri, te whakatakoto i te anga , te tuitui i ngā raraunga me ngā tauira, te tuku i tētahi mea e mahi ana, te ine pono, me te mahi anō me te kore raruraru. Ētahi tohu:

• Te whakamarie ki te porowhita katoa: raraunga → tauira → arotake → whakatakoto → aroturuki.

• Te whakaaro nui ki ngā whakamātautau tere, kaua ki ngā ariā taketake... me te nui o ngā ariā hei karo i ngā mahanga e kitea whānuitia ana.

• He putea mahi e whakaatu ana ka taea e koe te whakatutuki i ngā hua, ehara i te mea ko ngā pukatuhi anake.

• He whakaaro haepapa mō te mōrearea, te tūmataitinga, me te tika - ehara i te mahi, engari he mahi. Mā ngā kaupapa ahumahi pēnei i te Anga Whakahaere Mōrearea NIST AI me ngā Mātāpono OECD AI koe e āwhina ki te kōrero i te reo kotahi me ngā kaiarotake me ngā kaipupuri hea. [1][2]

He whaki iti: i ētahi wā ka tukuna atu e koe tētahi tauira, kātahi ka mōhio koe ko te kaupapa matua te wikitoria. He mea rerekē te ngākau mahaki.

He poto poto: i hangaia e tētahi tīma tētahi whakarōpūtanga ātaahua mō te whakarōpūtanga tautoko; i hinga i ngā ture kupumatua taketake i te wā whakautu tuatahi. I puritia e rātou ngā ture, i whakamahia te tauira mō ngā take taha, ā, i tukuna ngā mea e rua. He iti ake te makutu, he nui ake ngā putanga.

Te mahere ara mō te noho hei Kaiwhakawhanake AI 🗺️

Anei tētahi ara angiangi, auau hoki. Whakahokia kia maha ngā wā i a koe e piki ana i te taumata:

1. Te matatau ki te hōtaka i roto i te Python me ngā pukapuka matua o te DS: NumPy, pandas, scikit-learn. Pānuihia ngā aratohu mana, kātahi ka hanga tuhinga iti kia mōhio rawa ōu maihao. He pukapuka mahi whakamiharo te Aratohu Kaiwhakamahi . [3]

2. Ngā turanga ML mā roto i tētahi marautanga hanganga: ngā tauira rārangi, te whakatikatika, te whakamana whakawhiti, ngā inenga. He pai te mahi a ngā tuhipoka kauhau matarohia me te huinga akoranga tere mahi-ā-ringa.

3. Ngā taputapu ako hōhonu : whiriwhiria a PyTorch, a TensorFlow rānei, ā, akohia kia rite ki te whakangungu, te tiaki, me te uta i ngā tauira; te whakahaere i ngā huinga raraunga; me te whakatika i ngā hapa āhua noa. Tīmata ki ngā Akoranga PyTorch mēnā he pai ki a koe te "waehere tuatahi." [4]

4. Ngā kaupapa e tuku ana : te mōkihi me Docker, te oma aroturuki (ahakoa he rangitaki CSV kāore e pai ake), me te whakatinana i tētahi API iti rawa. Akohia a Kubernetes ina nui ake ngā whakatinanatanga pouaka-kotahi; ko Docker te tuatahi. [5]

5. Paparanga AI Kawenga : whakamahia he rārangi arowhai mōrearea māmā i ahu mai i te NIST/OECD (te mana, te pono, te mārama, te tika). Ka noho pūmau ngā kōrero, ā, ka noho hoha ngā arotake (i te huarahi pai). [1][2]

6. Whakamātautauhia he iti nei : NLP me ngā Transformers, tirohanga me ngā whakawhitiwhiti kōrero/ViT hou, ngā kaitohutohu, ngā taupānga LLM me ngā kaihoko rānei. Kōwhiria tētahi ara, hangaia kia rua ngā kaupapa iti, kātahi ka peka.

Ka hoki anō koe ki ngā taahiraa 2–6 ake tonu atu. He pono, koinei te mahi.

He puranga pūkenga ka whakamahia e koe i te nuinga o ngā rā 🧰

• Python + Raraunga tautohetohe : te tapahi i ngā rarangi, te hono, te whakarōpū, te whakapūtātanga. Ki te taea e koe te kanikani i ngā panda, ka māmā ake te whakangungu, ā, ka ma ake te aromatawai.

• Core ML : ngā wehewehenga tereina-whakamātautau, te karo i te turuturu, te mātauranga ine. Ko te aratohu scikit-learn tētahi o ngā tuhinga pai rawa atu mō te ara piki. [3]

• Anga DL : whiriwhiria tētahi, tīmatahia te mahi mai i te tīmatanga ki te mutunga, kātahi ka titiro ki tētahi atu i muri mai. Mā ngā tuhinga a PyTorch ka mārama te tauira hinengaro. [4]

• Te whakamātautau i te akuaku : ngā ara oma, ngā tautuhinga, me ngā taonga tuku iho. He mea whakarihariha ki a koe te whaipara tangata ā muri ake nei.

• Te whakapūmau me te whakarite : Docker hei tākai i tō puranga; Kubernetes ina hiahia koe ki ngā tārua, te tauine aunoa, me ngā whakahōutanga hurihuri. Tīmata ki konei. [5]

• Ngā kaupapa matua o te GPU : kia mōhio ai koe ki te wā tika hei reti i tētahi, me pēhea te pānga o te rahi o te puranga ki te putanga tukatuka, me te take e herea ai ētahi mahi ki te mahara.

• AI Kawenga : tuhia ngā pūtake raraunga, aromatawai i ngā mōrearea, me te whakamahere i ngā whakaiti mā te whakamahi i ngā āhuatanga mārama (te whai mana, te pono, te mārama, te tika). [1]

Marautanga tīmatanga: ngā hononga torutoru e kaha ake ana i tō rātou taumaha 🔗

• Ngā kaupapa ML : he huinga tuhipoka e kapi ana i ngā ariā me te akoranga tere. Honoa ēnei ki te mahi whakaharatau i roto i te ako scikit. [3]

• Ngā Anga : ngā Akoranga PyTorch (te Aratohu TensorFlow rānei mēnā he pai ki a koe a Keras). [4]

• Ngā mea nui o te pūtaiao raraunga te Aratohu Kaiwhakamahi a scikit-learn hei whakauru i ngā inenga, ngā tukanga, me te aromatawai. [3]

• Tukunga Tīmata a Docker kia huri ai te "mahi i runga i taku mīhini" ki te "mahi i ngā wāhi katoa." [5]

Tohua ēnei hei tohuwāhi. Ina raru koe, pānuihia tētahi whārangi, whakamātauria tētahi mea, tāruatia anō.

E toru ngā kaupapa putea e uiuitia ana 📁

1. Ngā whakautu pātai tāpiri-whakanui i runga i tō ake huinga raraunga

   • Waruhia/kawemaihia he pūtake mātauranga motuhake, hangaia he tāpiritanga me te tiki mai, tāpirihia he atanga kaiwhakamahi māmā.

   • Aroturuki i te whakaroa, te tika o tētahi huinga pātai me ngā whakautu kua whakaroahia, me ngā urupare a ngā kaiwhakamahi.

   • Whakaurua he wāhanga poto e kī ana "ngā take rahunga".

2. Tauira tirohanga me ngā herenga whakatinanatanga tūturu

   • Whakangunguhia he kaiwhakarōpū, he pūoko rānei, tuku mā te FastAPI, whakakīia ki roto i ngā ipu me te Docker, tuhia te āhua o tō tauine. [5]

   • Te kimi i te rerekētanga o ngā tuhinga (he tīmatanga pai te tatauranga taupori māmā i runga i ngā āhuatanga).

3. He rangahau take mō te AI kawenga

   • Kōwhiria he huinga raraunga tūmatanui me ngā āhuatanga tairongo. Whakaotia he tuhinga ine-me-te-whakaiti-i te āhua e hāngai ana ki ngā āhuatanga NIST (te mana, te pono, te tika). [1]

Me whai ia kaupapa i ēnei: ​​he README whārangi kotahi, he hoahoa, he tuhinga ka taea te tārua, me tētahi rārangi huringa iti. Tāpirihia he āhua emoji nā te mea, āe, ka pānuihia hoki ēnei e te tangata 🙂

Ngā MLOp, te whakatinanatanga, me te wāhanga kāore tetahi e ako i a koe 🚢

He pūkenga te tuku kaipuke. He iti noa te rere:

• Whakatakotoria tō taupānga ki roto i ngā ipu Docker kia dev ≈ prod. Tīmata ki ngā tuhinga Getting Started whaimana; neke ki Compose mō ngā tatūnga maha-ratonga. [5]

• Aroturuki i ngā whakamātautau (tae noa ki te rohe). Mā ngā tawhā, ngā inenga, ngā taonga tuku iho, me te tohu "toa" ka taea te mahi tahi me te pono o ngā tangohanga.

• Whakaritea ki a Kubernetes ina hiahia koe ki te tauine, ki te wehewehe rānei. Akohia ngā Whakatakotoranga, ngā Ratonga, me te whirihoranga whakapuakitanga i te tuatahi; ātete ki te hiahia ki te yak-shave.

• Ngā wā whakahaere kapua : Mahi tahi mō te hanga tauira; ngā tūāpapa whakahaeretia (SageMaker/Azure ML/Vertex) ina paahitia e koe ngā taupānga takaro.

• Te mōhiotanga ki te GPU : kāore koe e hiahia ki te tuhi i ngā kernel CUDA; me mōhio koe ki te wā e raru ai te uta raraunga.

He kupu whakarite iti nei he hapa: whakaarohia ngā MLOp he rite ki te tīmatanga rewena - whāngaihia ki te aunoatanga me te aroturukitanga, ki te kore ka haunga.

Ko te AI kawenga tō ara whakataetae 🛡️

Kei raro i te pēhanga ngā tīma kia whakaatu i te pono. Ki te taea e koe te kōrero pūmau mō te mōrearea, te tuhi tuhinga, me te whakahaere, ka noho koe hei tangata e hiahiatia ana e te iwi i roto i te rūma.

• Whakamahia he anga kua whakatūria : maherehia ngā whakaritenga ki ngā āhuatanga NIST (te mana, te pono, te mārama, te tika), kātahi ka hurihia hei tūemi rārangi arowhai me ngā paearu whakaae i roto i ngā PR. [1]

• Whakapūmau i ō mātāpono : e whakanui ana ngā Mātāpono AI o te OECD i ngā tika tangata me ngā uara manapori - he mea whai hua ina kōrerohia ngā taupatupatu. [2]

• Ngā tikanga matatika ngaio : he tohu poto ki tētahi waehere matatika i roto i ngā tuhinga hoahoa te rerekētanga i waenga i te "i whakaarohia e mātou" me te "i mahia noa e mātou".

Ehara tēnei i te mahi ture. He mahi toi tēnei.

Whakamātauria te mahi motuhake: whiriwhiria he ara, ā, akohia ōna taputapu 🛣️

• Ngā LLM me te NLP : ngā mahanga tohu, ngā matapihi horopaki, RAG, aromatawai i tua atu i te BLEU. Tīmata ki ngā paipa taumata-tiketike, kātahi ka whakarite.

• Te tirohanga whānui : te whakanui ake i ngā raraunga, te akuaku o ngā tapanga, me te whakatinanatanga ki ngā taputapu taha kei reira te whakaroa nui rawa atu.

• Ngā Kaitautoko : ngā āhuatanga rerekē o te urupare huna, ngā rautaki tīmatanga matao, me ngā KPI pakihi kāore e ōrite ki te RMSE.

• Ngā kaihoko me te whakamahinga o ngā taputapu : te karangatanga mahi, te wetewaehere here, me ngā ara haumaru.

He pono, whiriwhiria te rohe ipurangi e whakaohooho ana i a koe i ngā ata Rātapu.

Ripanga whakatairite: ngā huarahi mō te noho hei Kaiwhakawhanake AI 📊

Āe, he āhua koretake te tawhiti. Kāore i te tino tika ngā tēpu tuturu.

Ngā koropiko ako e piri tonu ana 🔁

• Ngā huringa e rua hāora te roa : 20 meneti te pānui tuhinga, 80 meneti te tuhi waehere, 20 meneti te tuhi i ngā mea i pakaru.

• He tuhinga kotahi whārangi : i muri i ia kaupapa poto, whakamāramahia te whakatakotoranga raruraru, ngā pūtake, ngā inenga, me ngā momo rahunga.

• Ngā herenga āta whakaarohia : whakangungua ki te CPU anake, kaua rānei he pukapuka ā-waho mō te tukatuka i mua, whakaritea rānei he tahua mō te 200 rārangi. Ka whakaputahia te auahatanga e ngā herenga, i tētahi ara.

• Ngā tere pepa : whakatinanahia te ngaronga raraunga, te uta raraunga rānei. Kāore koe e hiahia ki te SOTA hei ako nui.

Ki te kore e aro, he mea noa. Ka wiri te katoa. Haere, hoki mai, tukuna atu tetahi mea iti.

Te whakarite mō te uiuinga, engari kaua e whakamahia ngā whakaari 🎯

• Putea tuatahi : ngā repo tūturu i runga ake i ngā papa kiriata. Whakatakotoria kia kotahi te whakaaturanga iti.

• Whakamāramahia ngā taupatupatu : kia rite ki te tirotiro i ngā kōwhiringa inenga me pēhea te whakatika i tētahi hapa.

• Te whakaaro pūnaha : tuhia he raraunga → tauira → API → hoahoa aroturuki, kātahi ka kōrerotia.

• AI Kawenga : kia mau tonu he rārangi arowhai māmā e hangai ana ki te NIST AI RMF - he tohu tēnei mō te pakeketanga, ehara i te kupu rongonui. [1]

• Te matatau o te anga : whiriwhiria tētahi anga, ā, kia tūpono koe ki te whakamahi. He pai ngā tuhinga whaimana i roto i ngā uiuinga. [4]

Pukapuka tunu kai iti: tō kaupapa tuatahi mai i te mutunga ki te mutunga i te mutunga wiki 🍳

1. Raraunga : tīpakohia he huinga raraunga ma.

2. Pūtake : tauira ako-scikit me te whakamana-whakawhiti; takiuru i ngā inenga taketake. [3]

3. DL pass : mahi ōrite i roto i te PyTorch, i te TensorFlow rānei; whakatauritehia ngā aporo ki ngā aporo. [4]

4. Aroturuki : te oma o ngā rekoata (ahakoa he CSV māmā me ngā tohu wā). Tāpirihia te tohu ki te toa.

5. Mahi : tākaihia te matapae ki roto i te ara FastAPI, whakapūmau i te docker, whakahaere ā-rohe. [5]

6. Whakaarohia : he aha te inenga e whai tikanga ana ki te kaiwhakamahi, he aha ngā mōrearea kei reira, me ngā mea ka aroturukihia e koe i muri i te whakarewatanga - tangohia ngā kupu mai i te NIST AI RMF kia mau tonu ai te mārama. [1]

He tino pai tēnei? Kāore. He pai ake i te tatari mō te huarahi tino pai? Ae rawa.

Ngā hapa noa ka taea e koe te karo wawe ⚠️

• Te whakarahi rawa i tō ako ki ngā akoranga : he pai hei tīmatanga, engari me huri ki te whakaaro tuatahi ki te raruraru i ngā wā tata nei.

• Te peke i te hoahoa aromatawai : te tautuhi i te angitu i mua i te whakangungu. Ka penapena i ngā hāora.

• Te kore aro ki ngā kirimana raraunga : he nui ake ngā pūnaha ka pakaruhia e te nekehanga kaupapa i ngā tauira.

• Te wehi o te whakatinanatanga : He hoa ake a Docker i tōna āhua. Tīmata iti; whakaaetia he uaua te hanganga tuatahi. [5]

• Ka roa ngā tikanga matatika : ka whakapūmautia ā muri ake nei, ā, ka huri hei mahi ture. Whakauruhia kia rite ki te hoahoa - he māmā ake, he pai ake. [1][2]

Te TL;DR 🧡

Ki te maumahara koe ki tētahi mea: Ko te huarahi hei Kaiwhakawhanake AI ehara i te mea ko te kohikohi ariā, ko te whai rānei i ngā tauira kanapa. Engari ko te whakaoti rapanga tūturu me te whai i tētahi ara poto me te hinengaro haepapa. Akohia te puranga raraunga, whiriwhiria tētahi anga raraunga, tukuna ngā mea iti ki a Docker, aroturukihia āu mahi, ā, herea ō whiringa ki ngā aratohu whakaute pērā i a NIST me OECD. Hangaia kia toru ngā kaupapa iti, arohaina, ā, kōrerohia aua kaupapa pērā i te hoa mahi, kaua i te tohunga makutu. Koinā te nuinga - ko te nuinga.

Āe, me kī atu te kupu nei mēnā he āwhina: E mōhio ana ahau me pēhea te noho hei Kaiwhakawhanake AI . Kātahi ka whakamatau mā te kotahi hāora o te hanga arotahi i tēnei rā.

Ngā Tohutoro

[1] NIST. Anga Whakahaere Mōrearea o te Maramarama Horihori (AI RMF 1.0) . (PDF) - Hononga
[2] OECD. Ngā Mātāpono AI o te OECD - Tirohanga Whānui - Hononga
[3] scikit-learn. Aratohu Kaiwhakamahi (pūmau) - Hononga
[4] PyTorch. Ngā Akoranga (Akohia ngā Kaupapa Taketake, me ētahi atu) - Hononga
[5] Docker. Tīmata - Hononga