Ubuyobozi bwa tekiniki

Automatic Prompt Optimisation na DSPy

Automatic prompt optimization ikoresha algorithm mugushakisha amabwiriza meza cyangwa ingero nkeya.

Umva iki gitaboAI-byavuzwe · ~3 min

Kuriyi page4 min soma
  1. Incamake
  2. Kwibira cyane
  3. Gushyira mu bikorwa Isi
  4. Ejo hazaza ha Automatic Optimisation na DSPy
  5. Kubishyira mubikorwa
  6. Komeza Ubushakashatsi
  7. Ibibazo bikunze kubazwa

Incamake

Itanga amanota umukandida abajije ibipimo byerekana urugero aho kwishingikiriza kumagambo yahinduwe n'intoki. Uburyo buzwi cyane burimo APE, Google DeepMind ya OPRO na DSPy, isoko ifunguye kuva Stanford. Hamwe na hamwe bahindura injeniyeri yihuse kuva gukeka mubikorwa bipimwa, bisubirwamo.

Kwibira cyane

Intoki zateguwe n'intoki zifite intege nke zizwi. Gufata amajwi bikora kuri moderi imwe, dataset cyangwa verisiyo irashobora guhagarika bucece gukora kurindi, kandi impinduka zisanzwe zicirwa kuburorero buke gusa. Automatic prompt optimizasiyo isimbuza iyo loop hamwe nishakisha. Utanga ibintu bitatu: - Igikorwa. - Gushiraho inyongeramusaruro, nibyiza hamwe nibiteganijwe gusohoka. - Ibipimo, nkibihuye neza cyangwa rubricike yatsinzwe nubundi buryo. Optimizer itanga umukandida abaza, akabayobora, akabatsindira amanota kandi agakomeza abatsinze. Ingero zambere zatumye igitekerezo gifatika. APE. OPRO (Optimisation by PROmpting, Yang et al., Google DeepMind, 2023) yakoresheje LLM nka optimizer. Meta-prompt yayo yanditse urutonde rwambere hamwe namanota yabo, hamwe nicyitegererezo cyatanzwe gishya. OPRO yabonye amabwiriza nka "Fata umwuka uhagije kandi ukore kuri iki kibazo intambwe ku yindi," watsinze neza icyitegererezo kimwe ku mibare y'ishuri-shuri. Amagambo meza arashobora kuba adasanzwe kandi yihariye kurugero. DSPy igenda ihindura uburyo gahunda zanditswe. Yatunganijwe kuri Stanford NLP na Omar Khattab nabafatanyabikorwa kandi itangizwa mu 2023. Uratangaza icyo buri ntambwe ikora hamwe n'umukono, nka "ikibazo, imiterere -> igisubizo," hanyuma ugahuza modul nka Predict cyangwa ChainOfThought. Optimizer (mbere yitwa teleprompter), nka BootstrapFewShot cyangwa MIPROv2, hanyuma "ikusanya" porogaramu uhitamo imyigaragambyo n'amabwiriza kuri buri module kugirango wongere ibipimo byawe. Ibitekerezo bitatu bitari byo bikwiye gukosorwa: - Ibi bikoresho bikuraho ibitekerezo byabantu. Ntabwo babikora, kuko ibipimo n'ingero bisobanura "byiza." - Optimized prompts ifite umutekano mukwizera. Barashobora kurenga dataset nto. - Gukwirakwiza ni ubuntu. Igura guhamagara API nyayo, rimwe na rimwe.

Gushyira mu bikorwa Isi

Itsinda ryubaka inkunga-itike itondekanya yandika umukono wa DSPy "itike -> icyiciro" kandi itanga amatike 200 yanditseho na metero yukuri. Optimizer noneho atora imyigaragambyo mike-yerekana amanota meza.

Iyo isosiyete yimutse itanga LLM imwe ikajya mubindi, ikora optimizer ya DSPy kuri moderi nshya. Ntabwo yongeye kwandika intoki zahujwe na moderi ishaje.

Umushakashatsi akora loop-yuburyo bwa OPRO. Icyitegererezo kimwe gitanga amagambo mashya kubwinyigisho, buri jambo ryatsinzwe ku gipimo cyimibare, kandi amateka yatsindiye amanota asubira mubyiciro bikurikira.

Gusubiramo-byongerewe ibibazo-bisubiza umuyoboro uteganijwe kurangira kurangira, hamwe na metero igenzura ibisubizo byanyuma kubisubizo bifatika. Byombi kubaza-kwandika intambwe n'intambwe yo gusubiza itera imbere hamwe.

Ejo hazaza ha Automatic Optimisation na DSPy

Kwihutisha ibintu bigenda bihinduka igice gisanzwe cya LLM yubuhanga, hamwe na suite yo gusuzuma, aho kugira amatsiko yubushakashatsi. Haracyariho ibibazo bifunguye. Ni mu buhe buryo bwiza bushobora kwimurwa hagati yicyitegererezo? Nigute wandika ibipimo kumico ifatika nka tone cyangwa ubufasha? Nigute wakwirinda kurenza imibare mito? Ubushakashatsi bumwe bukomatanya guhita byoroha hamwe nu mucyo woroshye, kandi DSPy ishyigikira byombi. Abanyamideli bahinduka kenshi, kubwibyo ubushobozi bwo kongera gukora neza mu buryo bwikora kurwanya ibizamini bihamye birashoboka cyane kuruta ikintu icyo ari cyo cyose cyubwenge. Gusobanura ibipimo byiza n'ingero bikomeza kuba akazi ka muntu.

Kubishyira mubikorwa

Impamvu ari ngombwa

  • Igiciro na bije. Ibyemezo byubwubatsi bitwara imikorere nigiciro cyimikorere kumyaka.
  • Ibyemezo bisobanutse. Ubuhanga bwa tekinike bufasha amakipe guhitamo umurongo ukwiye, ntabwo ari shyashya gusa.
  • Kugenzura ubuziranenge. Guhitamo neza bya injeniyeri bigabanya ibintu byizewe mubikorwa.

Witondere

  • Gutezimbere igipimo kimwe gishobora guhisha intege nke za sisitemu.
  • Ibikorwa Remezo no kubungabunga akenshi usanga bidahabwa agaciro.
  • Icyuho cyumutekano no kwitegereza birashobora kwiyongera uko sisitemu igenda igorana.

Intambwe ikurikira

  1. Sobanura ubukererwe, ubuziranenge, nigiciro cyibiciro mbere yo kubishyira mubikorwa.
  2. Ibipimo byerekana umutwaro ufatika hamwe namakuru yimiterere.
  3. Gukurikirana ibikoresho kubikosa, drift, ningaruka zabakoresha.
  4. Tegura inzira yo gusubiza ibyabaye mbere yo gupima.

Komeza Ubushakashatsi

Ibaruwa y'amakuru y'ubuntu

Ibuka amakuru ya buri munsi ya AI

Inkuru eshatu zizewe za AI buri gitondo cy'iminsi y'akazi, zanditse mu rurimi rworoshye. Ubuntu iteka ryose, nta kwamamaza.

Imeri imwe buri munsi y'akazi. Kwiyimura mu buryo bumwe. Ntitwigera tugurisha cyangwa ngo dusangire aderesi yawe.

Isuzume

Take the Automatic Prompt Optimization and DSPy quiz

Ibisubizo ako kanya kuri buri kibazo, n'icyemezo gishobora gusangirwa gifite ID yagenzurwa umaze gutsinda isomo.

Tangira ikibazo

Fasha uburezi bwa AI bw'ubuntu. AI Understanding ni umuryango utegamiye kuri leta wa 501(c)(3) — nta kwamamaza, nta kwishyura, iteka ryose. Tanga impano

Ibibazo bikunze kubazwa

Niki Automatic Prompt Optimisation na DSPy?

Automatic prompt optimization ikoresha algorithm mugushakisha amabwiriza meza cyangwa ingero nkeya. Itanga amanota umukandida abajije ibipimo byerekana urugero aho kwishingikiriza kumagambo yahinduwe n'intoki. Uburyo buzwi cyane burimo APE, Google DeepMind ya OPRO na DSPy, isoko ifunguye kuva Stanford. Hamwe na hamwe bahindura injeniyeri yihuse kuva gukeka mubikorwa bipimwa, bisubirwamo.

Niki ugomba gutanga kugirango uhite ukora neza kugirango ukore?

Optimizer ikeneye umurimo, ingero na metric kugirango ibashe gutsinda amanota yumukandida kandi ikomeze ibyiza.

Muri OPRO, niki kigira uruhare rwa optimizer?

OPRO ikoresha icyitegererezo cyururimi nka optimizer. Irabona amabwiriza ashize hamwe n amanota yabo kandi igatanga andi mashya.

"OPRO uhumeka neza kandi ukore kuri iki kibazo intambwe ku yindi" byerekana iki?

Amabwiriza yatsindiye neza icyitegererezo kimwe kumibare-shuri. Gufata amajwi kubushakashatsi birashobora kuba bitangaje kandi ntibishobora kujyanwa mubindi byitegererezo.

Umukono wa DSPy ni iki?

Imikono itangaza icyo module igomba gukora mubijyanye ninjiza nibisohoka. DSPy noneho ikora ikibazo.

Nigute BootstrapFewShot ya DSPy ibona imyiyerekano yayo?

Inzira zitsindisha ziba ingero nkeya, harimo n'intambwe y'imbere utigeze wiyita wenyine.