{"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":"2026-08-12T18:21:41","modified_gmt":"2026-08-12T16:21:41","slug":"yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3","status":"publish","type":"post","link":"https:\/\/prohoster.info\/cs\/blog\/news\/yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3","title":{"rendered":"Jazykov\u00fd model s 28,9 milionu parametr\u016f pracuje na ESP32-S3","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

<\/i><\/a>
\nV\u00fdvoj\u00e1\u0159 Vja\u010deslav Serbov, vystupuj\u00edc\u00ed pod p\u0159ezd\u00edvkou slvDev, p\u0159edvedl<\/a> \u00fapln\u011b lok\u00e1ln\u00ed generaci textu na mikro\u0159adi\u010di ESP32-S3. Vytvo\u0159en\u00fd jazykov\u00fd model obsahuje 28,9 mili\u00f3nu parametr\u016f a vyd\u00e1v\u00e1 p\u0159ibli\u017en\u011b 9,9 tokenu za sekundu, ani\u017e by se obracel na server nebo jin\u00e9 extern\u00ed v\u00fdpo\u010detn\u00ed prost\u0159edky. Vygenerovan\u00e9 kr\u00e1tk\u00e9 p\u0159\u00edb\u011bhy se zobrazuj\u00ed na p\u0159ipojen\u00e9m OLED displeji.<\/p>\n

 <\/p>\n

Projekt esp32-ai<\/a> byl testov\u00e1n na modulu ESP32-S3 N16R8, kter\u00fd disponuje 512 KB vestav\u011bn\u00e9 SRAM, 8 MB PSRAM a 16 MB flash pam\u011bti. Kvantovan\u00e1 na \u010dty\u0159i bity model zab\u00edr\u00e1 14,9 MB. V repozit\u00e1\u0159i jsou zve\u0159ejn\u011bny zdrojov\u00e9 k\u00f3dy b\u011bhov\u00e9ho prost\u0159ed\u00ed v C, n\u00e1stroje pro u\u010den\u00ed a kvantov\u00e1n\u00ed modelu, firmware, testy a sc\u00e9n\u00e1\u0159e pro jeho nahr\u00e1v\u00e1n\u00ed na desku.<\/p>\n

 <\/p>\n

Uveden\u00fdch 28,9 mili\u00f3nu parametr\u016f by se nem\u011blo p\u0159\u00edmo srovn\u00e1vat s po\u010dtem parametr\u016f b\u011b\u017en\u00fdch desktopov\u00fdch LLM. Podle podrobn\u00e9 zpr\u00e1vy v\u00fdvoj\u00e1\u0159e<\/a>, model se skl\u00e1d\u00e1 z hust\u00e9ho v\u00fdpo\u010detn\u00edho j\u00e1dra p\u0159ibli\u017en\u011b na 559 tis\u00edc parametr\u016f, v\u00fdstupn\u00ed vrstvy na 3,1 milionu parametr\u016f a tabulky vrstevnat\u00fdch vektorov\u00fdch reprezentac\u00ed p\u0159ibli\u017en\u011b na 25 mili\u00f3n\u016f parametr\u016f. Pr\u00e1v\u011b posledn\u00ed zaji\u0161\u0165uje hlavn\u00ed \u010d\u00e1st form\u00e1ln\u00ed velikosti modelu, ale nen\u00ed zpracov\u00e1v\u00e1na cel\u00e1 p\u0159i generaci ka\u017ed\u00e9ho tokenu.<\/p>\n

 <\/p>\n

Pro um\u00edst\u011bn\u00ed modelu jsou vyu\u017eity t\u0159i \u00farovn\u011b pam\u011bti. Nej\u010dast\u011bji pou\u017e\u00edvan\u00e1 data se nach\u00e1zej\u00ed ve vnit\u0159n\u00ed SRAM, v\u00fdstupn\u00ed vrstva, KV-cache a do\u010dasn\u00e9 buffery \u2013 v PSRAM, zat\u00edmco 25-mili\u00f3nov\u00e1 tabulka z\u016fst\u00e1v\u00e1 ve flash pam\u011bti a je dostupn\u00e1 prost\u0159ednictv\u00edm mapov\u00e1n\u00ed do adresov\u00e9ho prostoru. Pro ka\u017ed\u00fd token je z n\u00ed na\u010d\u00edt\u00e1no jen p\u0159ibli\u017en\u011b \u0161est \u0159\u00e1dk\u016f s celkov\u00fdm objemem p\u0159ibli\u017en\u011b 450 bajt\u016f.<\/p>\n

 <\/p>\n

Takov\u00e9 za\u0159\u00edzen\u00ed je zalo\u017eeno na technologii Per-Layer Embeddings, kterou aplikoval Google v Gemma 3n. Podle dokumentace Google<\/a>, PLE-parametry mohou b\u00fdt ukl\u00e1d\u00e1ny mimo pracovn\u00ed pam\u011b\u0165 modelu a p\u0159id\u00e1v\u00e1ny do procesu v\u00fdstupu p\u0159i zpracov\u00e1n\u00ed jednotliv\u00fdch vrstev. V esp32-ai byl tento p\u0159\u00edstup p\u0159enesen na hierarchii pam\u011bti mikro\u0159adi\u010de, kde je rychl\u00e1 SRAM zvl\u00e1\u0161\u0165 omezen\u00e1.<\/p>\n

 <\/p>\n

Prvn\u00ed spr\u00e1vn\u00e1 verze prost\u0159ed\u00ed pro vykon\u00e1v\u00e1n\u00ed napsan\u00e9ho v C poskytovala pouze 0,57 tokenu za sekundu. Po um\u00edst\u011bn\u00ed v\u00fdstupn\u00ed vrstvy do PSRAM, p\u0159echodu na osmi bitov\u00e9 aktivace, rozd\u011blen\u00ed v\u00fdpo\u010dt\u016f mezi dv\u011b j\u00e1dra Xtensa LX7 a dal\u0161\u00edch optimalizac\u00edch byla latence sn\u00ed\u017eena na 94,9 ms na krok modelu. Kone\u010dn\u00e1 nam\u011b\u0159en\u00e1 rychlost dos\u00e1hla 9,88 tokenu za sekundu p\u0159i zohledn\u011bn\u00ed cel\u00e9ho procesu generov\u00e1n\u00ed. Hlavn\u00edm omezen\u00edm se nyn\u00ed stala propustnost PSRAM p\u0159i \u010dten\u00ed v\u00fdstupn\u00ed vrstvy.<\/p>\n

 <\/p>\n

Model byl natr\u00e9nov\u00e1n na syntetick\u00e9m datasetu TinyStories<\/a>, kter\u00fd se skl\u00e1d\u00e1 z jednoduch\u00fdch anglick\u00fdch p\u0159\u00edb\u011bh\u016f s omezen\u00fdm slovn\u00edkem. Proto je schopna pokra\u010dovat ve v\u011bt\u00e1ch a vytv\u00e1\u0159et mal\u00e9 souvisl\u00e9 p\u0159\u00edb\u011bhy, ale nen\u00ed ur\u010dena k odpov\u00edd\u00e1n\u00ed na ot\u00e1zky, prov\u00e1d\u011bn\u00ed pokyn\u016f, programov\u00e1n\u00ed nebo reprodukov\u00e1n\u00ed faktick\u00fdch znalost\u00ed. V\u00fdvoj\u00e1\u0159 pova\u017euje pr\u00e1ci p\u0159edev\u0161\u00edm za demonstraci efektivn\u00edho um\u00edst\u011bn\u00ed modelu do pam\u011bti mikrokontrol\u00e9ru, nikoli za hotov\u00e9ho autonomn\u00edho chatbotu.<\/p>\n

 <\/p>\n

V aktu\u00e1ln\u00ed dokumentaci autor zvl\u00e1\u0161\u0165 zd\u016fraz\u0148uje, \u017ee esp32-ai je samostatn\u00fd projekt. Projekty llama2.c<\/a> Andrey Karpathy a d\u0159\u00edv\u011bj\u0161\u00ed spu\u0161t\u011bn\u00ed 260tis\u00edcov\u00e9ho modelu na ESP32-S3 jsou uvedeny pouze jako orienta\u010dn\u00ed body a p\u0159\u00edklady p\u0159edchoz\u00edch prac\u00ed: k\u00f3d, z\u00e1chytn\u00e9 body a metodika nejsou z nich p\u0159\u00edmo p\u0159evzaty. Zdrojov\u00fd k\u00f3d esp32-ai se \u0161\u00ed\u0159\u00ed pod licenc\u00ed MIT<\/a>.<\/p>\n

 <\/p>\n

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

\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-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\udd47Jazykov\u00fd model s 28,9 miliony parametr\u016f fungoval na ESP32-S3 | ProHoster","description":".","canonical_url":"https:\/\/prohoster.info\/cs\/blog\/news\/yazykovaya-model-s-289-mln-parametrov-zarabotala-na-esp32-s3","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"cs_CZ","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:description":".","og:url":"https:\/\/prohoster.info\/cs\/blog\/news\/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":"2026-08-12T16:21:41+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"182824","title":null,"description":null,"keywords":null,"keyphrases":{"focus":[],"additional":[]},"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":"default","schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":null,"breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2026-08-11 06:43:49","updated":"2026-08-11 06:43:49","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/182824","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/comments?post=182824"}],"version-history":[{"count":1,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/182824\/revisions"}],"predecessor-version":[{"id":182897,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/182824\/revisions\/182897"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/media?parent=182824"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/categories?post=182824"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/tags?post=182824"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}