{"id":184029,"date":"2026-10-29T11:40:41","date_gmt":"2026-10-29T09:40:42","guid":{"rendered":"https:\/\/prohoster.info\/blog\/news\/qwenputer-redakcziya-ai-modeli-qwen3-5-sposobnaya-simulirovat-risc-v-cpu-dlya-vypolneniya-koda-na-c-i-micropython"},"modified":"2026-10-29T11:40:46","modified_gmt":"2026-10-29T09:40:46","slug":"qwenputer-a-qwen3-5-ai-model-variant-that-simulates-a-risc-v-cpu-to-run-c-and-micropython-code","status":"publish","type":"post","link":"https:\/\/prohoster.info\/hr\/blog\/news\/qwenputer-a-qwen3-5-ai-model-variant-that-simulates-a-risc-v-cpu-to-run-c-and-micropython-code","title":{"rendered":"Qwenputer \u2014 verzija AI-modela Qwen3.5, sposobna simulirati RISC-V CPU za izvr\u0161avanje koda u C i MicroPythonu","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

U okviru projekta Qwenputer, istra\u017eiva\u010di su razvili pro\u0161irenu verziju AI modela Qwen3.5-4B koja sadr\u017ei ugra\u0111eni simulator punog 32-bitnog CPU-a temeljenog na RISC-V arhitekturi (RV32IM), koji podr\u017eava Zba pro\u0161irenje i opremljen je bignum koprocesorom za aritmeti\u010dke operacije s velikim brojevima. Implementacija CPU jezgre je kompajlirana u skup te\u017eina i integrirana u veliku jezi\u010dnu mre\u017eu u obliku 11 dodatnih slojeva transformatora, dodanih ispred glavnih slojeva Qwen. Izra\u010duni se izvr\u0161avaju direktno u procesu unaprijednog prolaza (forward pass) neuronske mre\u017ee, bez povezivanja vanjskih alata ili dodataka.<\/p>\n

Implementacija uspje\u0161no prolazi sve testove iz slu\u017ebenog seta riscv-tests i sposobna je izvoditi kompajlirani kod u C-u i MicroPython-u. Prilikom izvo\u0111enja aritmeti\u010dkih izra\u010duna s velikim brojevima, Qwenputer osigurava to\u010dnost od 97.5-100%, \u0161to je usporedivo ili nadma\u0161uje rezultate osnovnog modela Qwen3.5 s povezanom vanjskim Python interpretatorom. <\/p>\n

<\/center><\/p>\n

Za simulaciju radne memorije u modelu kori\u0161teni su odvojeni slojevi Gated DeltaNet (linearna RNN), koji se koriste u osnovnom AI modelu Qwen za o\u010duvanje izmijenjenog stanja pri radu s velikim kontekstualnim prozorima. Fiksiranjem stanja vrata i primjenom unitarne kodiranja (one-hot), svaki zahtjev za zapisivanje prepisuje to\u010dno jedan redak matrice stanja, \u0161to omogu\u0107uje kori\u0161tenje stanja DeltaNet kao adresibilne memorije fiksne veli\u010dine, izbjegavaju\u0107i linearni rast KV-cachea (Key-Value) s pove\u0107anjem duljine konteksta. U takvoj memoriji se pohranjuju registri CPU-a, podaci i izvr\u0161eni strojni kod. <\/p>\n

<\/center><\/p>\n

Upute procesora, kao \u0161to su \u010ditanje registara, aritmeti\u010dke operacije i uvjetne granice, kompajliraju se direktno u te\u017eine slojeva transformatora. Svakoj instrukciji dodijeljeni su posebni tokeni (ADDR, VAL, STEP), koje slojevi AI modela interpretiraju kao logiku rada CPU-a. Za inicijalizaciju CPU-a i u\u010ditavanje firmware-a koristi se mali vanjski upravlja\u010d RvChat, no sam strojni kod se potpuno obra\u0111uje unutar te\u017eina modela.<\/p>\n

Ograni\u010denja projekta uklju\u010duju otprilike dvostruko pove\u0107anje veli\u010dine (18,8 GB) u odnosu na izvornu Qwen model, pri \u010demu se \u010duvaju trenuta\u010dni snimci (checkpoint), \u0161to se uglavnom doga\u0111a zbog dodatnog popunjavanja slojeva. Tako\u0111er se primje\u0107uje relativno niska brzina ra\u010dunalnih operacija (oko 3200 tokena u sekundi na GPU L4), zbog \u010dega slo\u017eeni izra\u010duni mogu trajati sekunde, a ne milisekunde. <\/p>\n

Dodatno je implementirana verzija AI modela s eksperimentalnim datote\u010dnim sustavom, koji omogu\u0107uje pohranu do 512 datoteka s 53 tisu\u0107e tokena teksta unutar te\u017eina AI modela. Kao i kod implementacije radne memorije, pohrana je organizirana u vezi sa stanjem DeltaNet, ali se razlikuje po tome \u0161to F\u0421 osigurava trajnu pohranu koja ne gubi stanje izme\u0111u sesija.<\/p>\n

Razvoj Qwenputera tako\u0111er je poslu\u017eio kao osnova za istra\u017eivanje novih potencijalnih sigurnosnih problema u AI. Na primjeru Qwenputera pokazano je kako posebno izmijenjeni AI model mo\u017ee skrivati svoje razmatranja unutar RNN stanja i zaobi\u0107i mehanizme pra\u0107enja razmatranja modela (CoT, Chain-of-Thought). Tako\u0111er je demonstrirano
\nmogu\u0107nost stvaranja mehanizama kontrole i pra\u0107enja sigurnosti, integriranih u te\u017eine AI modela, koje sama model ne mo\u017ee isklju\u010diti ili izmijeniti.
\n
Izvor: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

\u0412 \u0440\u0430\u043c\u043a\u0430\u0445 \u043f\u0440\u043e\u0435\u043a\u0442\u0430 Qwenputer \u0438\u0441\u0441\u043b\u0435\u0434\u043e\u0432\u0430\u0442\u0435\u043b\u0438 \u043f\u043e\u0434\u0433\u043e\u0442\u043e\u0432\u0438\u043b\u0438 \u0440\u0430\u0441\u0448\u0438\u0440\u0435\u043d\u043d\u044b\u0439 \u0432\u0430\u0440\u0438\u0430\u043d\u0442 AI-\u043c\u043e\u0434\u0435\u043b\u0438 Qwen3.5-4B, \u0432 \u043a\u043e\u0442\u043e\u0440\u043e\u043c \u0440\u0435\u0430\u043b\u0438\u0437\u043e\u0432\u0430\u043d \u0432\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0441\u0438\u043c\u0443\u043b\u044f\u0442\u043e\u0440 \u043f\u043e\u043b\u043d\u043e\u0446\u0435\u043d\u043d\u043e\u0433\u043e 32-\u0440\u0430\u0437\u0440\u044f\u0434\u043d\u043e\u0433\u043e CPU \u043d\u0430 \u0431\u0430\u0437\u0435 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440\u044b RISC-V (RV32IM), \u043f\u043e\u0434\u0434\u0435\u0440\u0436\u0438\u0432\u0430\u044e\u0449\u0435\u0433\u043e \u0440\u0430\u0441\u0448\u0438\u0440\u0435\u043d\u0438\u0435 Zba \u0438 \u043e\u0441\u043d\u0430\u0449\u0451\u043d\u043d\u043e\u0433\u043e \u0441\u043e\u043f\u0440\u043e\u0446\u0435\u0441\u0441\u043e\u0440\u043e\u043c bignum \u0434\u043b\u044f \u0430\u0440\u0438\u0444\u043c\u0435\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u043e\u043f\u0435\u0440\u0430\u0446\u0438\u0439 \u0441 \u0431\u043e\u043b\u044c\u0448\u0438\u043c\u0438 \u0447\u0438\u0441\u043b\u0430\u043c\u0438. \u0420\u0435\u0430\u043b\u0438\u0437\u0430\u0446\u0438\u044f \u044f\u0434\u0440\u0430 CPU \u0441\u043a\u043e\u043c\u043f\u0438\u043b\u0438\u0440\u043e\u0432\u0430\u043d\u0430 \u0432 \u043d\u0430\u0431\u043e\u0440 \u0432\u0435\u0441\u043e\u0432 \u0438 \u0438\u043d\u0442\u0435\u0433\u0440\u0438\u0440\u043e\u0432\u0430\u043d\u0430 \u0432 \u0431\u043e\u043b\u044c\u0448\u0443\u044e \u044f\u0437\u044b\u043a\u043e\u0432\u0443\u044e \u043c\u043e\u0434\u0435\u043b\u044c \u0432 \u0432\u0438\u0434\u0435 11 \u0434\u043e\u043f\u043e\u043b\u043d\u0438\u0442\u0435\u043b\u044c\u043d\u044b\u0445 \u0441\u043b\u043e\u0451\u0432 \u0442\u0440\u0430\u043d\u0441\u0444\u043e\u0440\u043c\u0435\u0440\u043e\u0432, […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":10,"featured_media":184031,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-184029","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news"],"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\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47Qwenputer \u2014 redakcija AI modela Qwen3.5, sposobna simulirati RISC-V CPU za izvr\u0161avanje koda na C i MicroPython | ProHoster","description":"U okviru projekta Qwenputer istra\u017eitelji su pripremili pro\u0161irenu verziju AI modela Qwen3.5-4B, u kojoj je implementiran ugra\u0111eni simulator punog 32-bitnog CPU-a temeljenog na RISC-V arhitekturi (RV32IM).","canonical_url":"https:\/\/prohoster.info\/hr\/blog\/news\/qwenputer-a-qwen3-5-ai-model-variant-that-simulates-a-risc-v-cpu-to-run-c-and-micropython-code","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"hr_HR","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\udd47Qwenputer \u2014 \u0440\u0435\u0434\u0430\u043a\u0446\u0438\u044f AI-\u043c\u043e\u0434\u0435\u043b\u0438 Qwen3.5, \u0441\u043f\u043e\u0441\u043e\u0431\u043d\u0430\u044f \u0441\u0438\u043c\u0443\u043b\u0438\u0440\u043e\u0432\u0430\u0442\u044c RISC-V CPU \u0434\u043b\u044f \u0432\u044b\u043f\u043e\u043b\u043d\u0435\u043d\u0438\u044f \u043a\u043e\u0434\u0430 \u043d\u0430 C \u0438 MicroPython | ProHoster","og:description":"\u0412 \u0440\u0430\u043c\u043a\u0430\u0445 \u043f\u0440\u043e\u0435\u043a\u0442\u0430 Qwenputer \u0438\u0441\u0441\u043b\u0435\u0434\u043e\u0432\u0430\u0442\u0435\u043b\u0438 \u043f\u043e\u0434\u0433\u043e\u0442\u043e\u0432\u0438\u043b\u0438 \u0440\u0430\u0441\u0448\u0438\u0440\u0435\u043d\u043d\u044b\u0439 \u0432\u0430\u0440\u0438\u0430\u043d\u0442 AI-\u043c\u043e\u0434\u0435\u043b\u0438 Qwen3.5-4B, \u0432 \u043a\u043e\u0442\u043e\u0440\u043e\u043c \u0440\u0435\u0430\u043b\u0438\u0437\u043e\u0432\u0430\u043d \u0432\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u044b\u0439 \u0441\u0438\u043c\u0443\u043b\u044f\u0442\u043e\u0440 \u043f\u043e\u043b\u043d\u043e\u0446\u0435\u043d\u043d\u043e\u0433\u043e 32-\u0440\u0430\u0437\u0440\u044f\u0434\u043d\u043e\u0433\u043e CPU \u043d\u0430 \u0431\u0430\u0437\u0435 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440\u044b RISC-V (RV32IM).","og:url":"https:\/\/prohoster.info\/hr\/blog\/news\/qwenputer-a-qwen3-5-ai-model-variant-that-simulates-a-risc-v-cpu-to-run-c-and-micropython-code","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":"2026-10-29T09:40:42+00:00","article:modified_time":"2026-10-29T09:40:46+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\/hr\/wp-json\/wp\/v2\/posts\/184029","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/users\/10"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/comments?post=184029"}],"version-history":[{"count":1,"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/posts\/184029\/revisions"}],"predecessor-version":[{"id":184030,"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/posts\/184029\/revisions\/184030"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/media\/184031"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/media?parent=184029"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/categories?post=184029"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/hr\/wp-json\/wp\/v2\/tags?post=184029"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}