{"id":107246,"date":"2023-03-13T18:48:30","date_gmt":"2023-03-13T16:48:30","guid":{"rendered":"https:\/\/prohoster.info\/?p=107246"},"modified":"2023-03-14T14:08:05","modified_gmt":"2023-03-14T12:08:05","slug":"opublikovan-openchatkit-instrumentarij-dlya-sozdaniya-chatbotov","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/opublikovan-openchatkit-instrumentarij-dlya-sozdaniya-chatbotov","title":{"rendered":"OpenChatKit \u00ebsht\u00eb publikuar, nj\u00eb mjete p\u00ebr krijimin e chatbot-\u00ebve.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
OpenChatKit, nj\u00eb mjet i hapur, ka si q\u00ebllim thjeshtimin e krijimit t\u00eb chatbot-\u00ebve p\u00ebr p\u00ebrdorim t\u00eb specializuar dhe t\u00eb p\u00ebrgjithsh\u00ebm. Sistemi \u00ebsht\u00eb i adaptruar p\u00ebr t\u00eb realizuar detyra t\u00eb tilla si p\u00ebrgjigjja n\u00eb pyetje, mbajtja e bisedave shum\u00eb p\u00ebrb\u00ebr\u00eb, p\u00ebrmbledhja, nxjerrja e informacionit dhe klasifikimi i tekstit. Kodi \u00ebsht\u00eb shkruar n\u00eb gjuh\u00ebn Python dhe shp\u00ebrndahet n\u00ebn licenc\u00ebn Apache 2.0. Projekti p\u00ebrfshin nj\u00eb model t\u00eb gatsh\u00ebm, kod p\u00ebr trajnim t\u00eb modelit tuaj, mjete p\u00ebr testimin e rezultateve t\u00eb modelit, mjete p\u00ebr plot\u00ebsimin e modelit me kontekst nga nj\u00eb indeks i jasht\u00ebm dhe adaptron modelin baz\u00eb p\u00ebr zgjidhjen e detyrave tuaja. <\/p>\n
Si baz\u00eb p\u00ebr botin, ofrohet nj\u00eb model i thjesht\u00eb i m\u00ebsimit t\u00eb makinerive (GPT-NeoXT-Chat-Base-20B), i nd\u00ebrtuar me nj\u00eb model gjuh\u00ebsor q\u00eb mbulon rreth 20 miliard parametra dhe i optimizuar p\u00ebr komunikim konverzativ. P\u00ebr trajnimin e modelit jan\u00eb p\u00ebrdorur t\u00eb dh\u00ebna nga koleksionet e projekteve LAION, Together dhe Ontocord.ai. <\/p>\n
P\u00ebr t\u00eb zgjeruar baz\u00ebn ekzistuese t\u00eb njohurive, ofrohet nj\u00eb sistem q\u00eb mund t\u00eb nxjerr\u00eb informacion t\u00eb shtuar nga depo t\u00eb jashtme, API dhe burime t\u00eb tjera. P\u00ebr shembull, \u00ebsht\u00eb e mundur p\u00ebrdit\u00ebsimi i informacionit p\u00ebrmes t\u00eb dh\u00ebnave nga Wikipedia dhe lajmet. S\u00eb ngritur, \u00ebsht\u00eb n\u00eb dispozicion nj\u00eb model p\u00ebr moderim, i st\u00ebrvitur me p\u00ebrdorimin e 6 miliard parametrave, i bazuar n\u00eb modelin GPT-JT dhe i destinuar p\u00ebr t\u00eb filtruar pyetje t\u00eb papranueshme ose p\u00ebr t\u00eb kufizuar diskutimet brenda temave t\u00eb caktuara. <\/p>\n
Ve\u00e7mas mund t\u00eb p\u00ebrmendet projekti ChatLLaMA, i cili ofron nj\u00eb bibliotek\u00eb p\u00ebr krijimin e ndihm\u00ebsve t\u00eb zgjuar, t\u00eb ngjash\u00ebm me ChatGPT. Projekti zhvillohet me synimin p\u00ebr t'u ekzekutuar n\u00eb pajisjet e veta dhe p\u00ebr t\u00eb krijuar zgjidhje t\u00eb personalizuara, t\u00eb dizajnuara p\u00ebr t\u00eb p\u00ebrfshir\u00eb fusha t\u00eb ngushta t\u00eb njohurive (p\u00ebr shembull, mjek\u00ebsi, jurisprudenc\u00eb, lojra, k\u00ebrkime shkencore, etj.). Kodi i ChatLLaMA ofrohet n\u00ebn licenc\u00ebn GPLv3. <\/p>\n
Projekti mb\u00ebshtet p\u00ebrdorimin e modeleve t\u00eb bazuara n\u00eb arkitektur\u00ebn LLaMA (Large Language Model Meta AI), e propozuar nga kompania Meta. Modeli i plot\u00eb i LLaMA mbulon 65 miliard\u00eb parametra, por p\u00ebr ChatLLaMA rekomandohet t\u00eb p\u00ebrdoren variante me 7 dhe 13 miliard\u00eb parametra ose modelet GPTJ (6 miliard), GPTNeoX (1.3 miliard), 20BOPT (13 miliard), BLOOM (7.1 miliard) dhe Galactica (6.7 miliard). Fillimisht, modelet LLaMA ofrohen vet\u00ebm p\u00ebr k\u00ebrkuesit me k\u00ebrkes\u00eb t\u00eb ve\u00e7ant\u00eb, por pasi t\u00eb dh\u00ebnat u shp\u00ebrndan\u00eb p\u00ebrmes torent\u00ebve, entuziast\u00ebt p\u00ebrgatit\u00ebn nj\u00eb skript q\u00eb lejon shkarkimin e modelit nga kushdo q\u00eb e d\u00ebshiron. \n Burimi: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"