{"id":182824,"date":"2026-08-11T04:54:15","date_gmt":"2026-08-11T02:54:15","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3"},"modified":"2026-08-11T04:54:15","modified_gmt":"2026-08-11T02:54:15","slug":"yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3","title":{"rendered":"Modeli gjuh\u00ebsor me 28.9 milion parametra ka filluar t\u00eb funksionoj\u00eb n\u00eb ESP32-S3","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

<\/i><\/a>
1 <\/p>\n

Zhvilluesi Vyacheslav Serbov, q\u00eb vepron me emrin e p\u00ebrdorimit slvDev, demonstroi<\/a> gjen rrug\u00ebn e plot\u00eb t\u00eb gjenerimit t\u00eb tekstit n\u00eb mikro-kontrollerin ESP32-S3. Modeli gjuh\u00ebsor i krijuar nga ai p\u00ebrmban 28,9 milion parametra dhe jep rreth 9,9 tokene n\u00eb sekond\u00eb, pa u drejtuar n\u00eb server apo burime t\u00eb tjera t\u00eb jashtme t\u00eb llogaritjes. Tregimet e shkurtra t\u00eb gjeneruara shfaqen n\u00eb ekranin OLED t\u00eb lidhur.<\/p>\n

<\/noindex> <\/p>\n

Projekt esp32-ai<\/a> \u00ebsht\u00eb testuar n\u00eb modulin ESP32-S3 N16R8, i cili ka 512 KB SRAM t\u00eb brendshme, 8 MB PSRAM dhe 16 MB memorie flash. Modeli i kuantizuar deri n\u00eb kat\u00ebr bit z\u00eb 14,9 MB. N\u00eb depo jan\u00eb publikuar kodin burimor t\u00eb mjedisit t\u00eb ekzekutimit n\u00eb C, mjetet e trajnimit dhe kuantizimit t\u00eb modelit, firmware, testet dhe skenar\u00ebt e ngarkimit t\u00eb tij n\u00eb bord.<\/p>\n

<\/noindex> <\/p>\n

T\u00eb deklaruar 28,9 milion parametra nuk duhet t\u00eb krahasohen drejtp\u00ebrdrejt me numrin e parametrave t\u00eb LLM-ve t\u00eb zakonshme desktop. Sipas raportit t\u00eb detajuar t\u00eb zhvilluesit<\/a>, modeli p\u00ebrb\u00ebhet nga nj\u00eb b\u00ebrtham\u00eb t\u00eb dendur llogarit\u00ebse me rreth 559 mij\u00eb parametra, nj\u00eb shtres\u00eb dal\u00ebse me 3,1 milion parametra dhe nj\u00eb tabel\u00eb p\u00ebrfaq\u00ebsuese vektor\u00ebsh p\u00ebr \u00e7do shtres\u00eb me rreth 25 milion parametra. Pik\u00ebrisht kjo e fundit ofron pjes\u00ebn kryesore t\u00eb madh\u00ebsis\u00eb formale t\u00eb modelit, por nuk trajtohet e t\u00ebra gjat\u00eb gjenerimit t\u00eb \u00e7do tokeni.<\/p>\n

<\/noindex> <\/p>\n

P\u00ebr t\u00eb akomoduar modelin, jan\u00eb aktivizuar tre nivele memoriesh. T\u00eb dh\u00ebnat m\u00eb t\u00eb p\u00ebrdorura ndodhen n\u00eb SRAM-in e brendsh\u00ebm, shtresa dal\u00ebse, KV-cache dhe buferat e p\u00ebrkohsh\u00ebm - n\u00eb PSRAM, nd\u00ebrsa tabela me 25 milion ndodhet n\u00eb memorien flash dhe \u00ebsht\u00eb e aksesueshme p\u00ebrmes hartimit n\u00eb hap\u00ebsir\u00ebn adresuese. P\u00ebr \u00e7do token lexohen vet\u00ebm rreth gjasht\u00eb rreshta me nj\u00eb v\u00ebllim total prej rreth 450 byte.<\/p>\n

<\/p>\n

Nj\u00eb pajisje e till\u00eb bazohet n\u00eb teknologjin\u00eb Per-Layer Embeddings, e aplikuar nga Google n\u00eb Gemma 3n. Sipas dokumentacionit t\u00eb Google<\/a>, parametrat PLE mund t\u00eb ruhen jasht\u00eb memories operative t\u00eb modelit dhe t\u00eb shtohen n\u00eb procesin e daljes nd\u00ebrsa p\u00ebrpunohen shtresa t\u00eb ve\u00e7anta. N\u00eb esp32-ai ky qasje \u00ebsht\u00eb transferuar n\u00eb hierarkin\u00eb e memories s\u00eb mikro-kontrollerit, ku SRAM-i i shpejt\u00eb \u00ebsht\u00eb ve\u00e7an\u00ebrisht i kufizuar.<\/p>\n

<\/noindex> <\/p>\n

Versioni e par\u00eb e sakt\u00eb e ambientit t\u00eb ekzekutimit t\u00eb shkruar n\u00eb C jepte vet\u00ebm 0,57 tokena n\u00eb sekond\u00eb. Pasi vendos\u00ebm shtres\u00ebn e daljes n\u00eb PSRAM, kaluam n\u00eb aktivizime me tet\u00eb bit\u00eb, shp\u00ebrndam\u00eb llogaritjet midis dy b\u00ebrthameve Xtensa LX7 dhe kryem optimizime t\u00eb tjera, vonesa u ul n\u00eb 94,9 ms p\u00ebr hapin e modelit. Shpejt\u00ebsia e matur p\u00ebrfundimtar arriti n\u00eb 9,88 tokena n\u00eb sekond\u00eb duke marr\u00eb parasysh procesin e plot\u00eb t\u00eb gjenerimit. Kufiri kryesor tani \u00ebsht\u00eb kapaciteti i PSRAM p\u00ebr leximin e shtres\u00ebs s\u00eb daljes.<\/p>\n

<\/p>\n

Modeli \u00ebsht\u00eb trajnuar n\u00eb nj\u00eb grup t\u00eb dh\u00ebnash sintetik TinyStories<\/a>, i p\u00ebrb\u00ebr\u00eb nga tregime t\u00eb thjeshta n\u00eb anglisht me nj\u00eb fjalor t\u00eb kufizuar. Prandaj, ajo \u00ebsht\u00eb n\u00eb gjendje t\u00eb vazhdoj\u00eb frazat dhe t\u00eb krijoj\u00eb histori t\u00eb vogla koherente, por nuk \u00ebsht\u00eb e destinuar p\u00ebr t\u00eb dh\u00ebn\u00eb p\u00ebrgjigje n\u00eb pyetje, p\u00ebr t\u00eb kryer udh\u00ebzime, p\u00ebr programim ose p\u00ebr t\u00eb rikujtuar njohuri faktike. Zhvilluesi e sheh pun\u00ebn kryesisht si nj\u00eb demonstrim t\u00eb vendosjes efektive t\u00eb modelit n\u00eb kujtes\u00ebn e mikrocontroller-it, e jo si nj\u00eb chatbot autonom t\u00eb gatsh\u00ebm.<\/p>\n

<\/noindex> <\/p>\n

N\u00eb dokumentacionin aktual, autori thekson ve\u00e7mas se esp32-ai \u00ebsht\u00eb nj\u00eb zhvillim i pavarur. Projektet llama2.c<\/a> t\u00eb Andrey Karpaty dhe fillimi m\u00eb her\u00ebt i modelit me 260,000 parametra n\u00eb ESP32-S3 p\u00ebrmenden vet\u00ebm si orientime dhe shembuj t\u00eb aktiviteteve t\u00eb m\u00ebparshme: kodi, pikat e kontrollit dhe metoda nuk jan\u00eb marr\u00eb drejtp\u00ebrdrejt nga ato. Kodi burimor i esp32-ai distribuohet n\u00ebn licenc\u00ebn MIT<\/a>.<\/p>\n

<\/noindex><\/p>\n

Burimi: linux.org.ru<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

1 \u0420\u0430\u0437\u0440\u0430\u0431\u043e\u0442\u0447\u0438\u043a \u0412\u044f\u0447\u0435\u0441\u043b\u0430\u0432 \u0421\u0435\u0440\u0431\u043e\u0432, \u0432\u044b\u0441\u0442\u0443\u043f\u0430\u044e\u0449\u0438\u0439 \u043f\u043e\u0434 \u043f\u0441\u0435\u0432\u0434\u043e\u043d\u0438\u043c\u043e\u043c slvDev, \u043f\u0440\u043e\u0434\u0435\u043c\u043e\u043d\u0441\u0442\u0440\u0438\u0440\u043e\u0432\u0430\u043b \u043f\u043e\u043b\u043d\u043e\u0441\u0442\u044c\u044e \u043b\u043e\u043a\u0430\u043b\u044c\u043d\u0443\u044e \u0433\u0435\u043d\u0435\u0440\u0430\u0446\u0438\u044e \u0442\u0435\u043a\u0441\u0442\u0430 \u043d\u0430 \u043c\u0438\u043a\u0440\u043e\u043a\u043e\u043d\u0442\u0440\u043e\u043b\u043b\u0435\u0440\u0435 ESP32-S3. \u0421\u043e\u0437\u0434\u0430\u043d\u043d\u0430\u044f \u0438\u043c \u044f\u0437\u044b\u043a\u043e\u0432\u0430\u044f \u043c\u043e\u0434\u0435\u043b\u044c \u0441\u043e\u0434\u0435\u0440\u0436\u0438\u0442 28,9 \u043c\u043b\u043d \u043f\u0430\u0440\u0430\u043c\u0435\u0442\u0440\u043e\u0432 \u0438 \u0432\u044b\u0434\u0430\u0451\u0442 \u043e\u043a\u043e\u043b\u043e 9,9 \u0442\u043e\u043a\u0435\u043d\u0430 \u0432 \u0441\u0435\u043a\u0443\u043d\u0434\u0443, \u043d\u0435 \u043e\u0431\u0440\u0430\u0449\u0430\u044f\u0441\u044c \u043a \u0441\u0435\u0440\u0432\u0435\u0440\u0443 \u0438\u043b\u0438 \u0434\u0440\u0443\u0433\u0438\u043c \u0432\u043d\u0435\u0448\u043d\u0438\u043c \u0432\u044b\u0447\u0438\u0441\u043b\u0438\u0442\u0435\u043b\u044c\u043d\u044b\u043c \u0440\u0435\u0441\u0443\u0440\u0441\u0430\u043c. \u0421\u0433\u0435\u043d\u0435\u0440\u0438\u0440\u043e\u0432\u0430\u043d\u043d\u044b\u0435 \u043a\u043e\u0440\u043e\u0442\u043a\u0438\u0435 \u0440\u0430\u0441\u0441\u043a\u0430\u0437\u044b \u0432\u044b\u0432\u043e\u0434\u044f\u0442\u0441\u044f \u043d\u0430 \u043f\u043e\u0434\u043a\u043b\u044e\u0447\u0451\u043d\u043d\u044b\u0439 OLED-\u0434\u0438\u0441\u043f\u043b\u0435\u0439. \u041f\u0440\u043e\u0435\u043a\u0442 esp32-ai \u0438\u0441\u043f\u044b\u0442\u0430\u043d \u043d\u0430 \u043c\u043e\u0434\u0443\u043b\u0435 ESP32-S3 N16R8, \u0440\u0430\u0441\u043f\u043e\u043b\u0430\u0433\u0430\u044e\u0449\u0435\u043c […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":9,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-182824","post","type-post","status-publish","format-standard","hentry","category-novosti-interneta"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47Modeli gjuh\u00ebsor me 28,9 milion parametra funksionon n\u00eb ESP32-S3 | ProHoster","description":"","canonical_url":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"sq_AL","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47\u042f\u0437\u044b\u043a\u043e\u0432\u0430\u044f \u043c\u043e\u0434\u0435\u043b\u044c \u0441 28,9 \u043c\u043b\u043d \u043f\u0430\u0440\u0430\u043c\u0435\u0442\u0440\u043e\u0432 \u0437\u0430\u0440\u0430\u0431\u043e\u0442\u0430\u043b\u0430 \u043d\u0430 ESP32-S3 | ProHoster","og:url":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2026-08-11T02:54:15+00:00","article:modified_time":"2026-08-11T02:54:15+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":[],"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts\/182824","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/comments?post=182824"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts\/182824\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/media?parent=182824"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/categories?post=182824"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/tags?post=182824"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}