{"id":251496,"date":"2024-09-25T05:30:00","date_gmt":"2024-09-25T05:30:00","guid":{"rendered":"https:\/\/datascientest.com\/?p=251496"},"modified":"2024-10-03T11:01:54","modified_gmt":"2024-10-03T11:01:54","slug":"llama-tout-savoir","status":"publish","type":"post","link":"https:\/\/datascientest.com\/llama-tout-savoir","title":{"rendered":"LLaMa : tout savoir sur le mod\u00e8le de langage de Meta"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"251496\" class=\"elementor elementor-251496\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-a1b23fd e-flex e-con-boxed e-con\" data-id=\"a1b23fd\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-5c11447 elementor-widget elementor-widget-heading\" data-id=\"5c11447\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-heading-title{padding:0;margin:0;line-height:1}.elementor-widget-heading .elementor-heading-title[class*=elementor-size-]>a{color:inherit;font-size:inherit;line-height:inherit}.elementor-widget-heading .elementor-heading-title.elementor-size-small{font-size:15px}.elementor-widget-heading .elementor-heading-title.elementor-size-medium{font-size:19px}.elementor-widget-heading .elementor-heading-title.elementor-size-large{font-size:29px}.elementor-widget-heading .elementor-heading-title.elementor-size-xl{font-size:39px}.elementor-widget-heading .elementor-heading-title.elementor-size-xxl{font-size:59px}<\/style><p class=\"elementor-heading-title elementor-size-default\">Contre toute attente, LLaMa, le mod\u00e8le d\u2019intelligence artificielle propos\u00e9 par Meta, conna\u00eet un succ\u00e8s remarquable. Au risque de faire de l\u2019ombre \u00e0 ChatGPT, Google Gemini et autres stars du lot. Comment expliquer cette mont\u00e9e en popularit\u00e9 rapide et soudaine ?<\/p>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-c591fd4 elementor-widget elementor-widget-text-editor\" data-id=\"c591fd4\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-text-editor.elementor-drop-cap-view-stacked .elementor-drop-cap{background-color:#69727d;color:#fff}.elementor-widget-text-editor.elementor-drop-cap-view-framed .elementor-drop-cap{color:#69727d;border:3px solid;background-color:transparent}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap{margin-top:8px}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap-letter{width:1em;height:1em}.elementor-widget-text-editor .elementor-drop-cap{float:left;text-align:center;line-height:1;font-size:50px}.elementor-widget-text-editor .elementor-drop-cap-letter{display:inline-block}<\/style>\t\t\t\t<p><span style=\"font-weight: 400;\">Pour un succ\u00e8s, c\u2019est un succ\u00e8s. Fin ao\u00fbt 2024, LLaMa d\u00e9nombrait 350 millions de t\u00e9l\u00e9chargements, pulv\u00e9risant les records en mati\u00e8re de <a href=\"https:\/\/datascientest.com\/large-language-models-tout-savoir\">LLM (Large Langage Model<\/a> \u2013 un mod\u00e8le con\u00e7u pour comprendre et g\u00e9n\u00e9rer du texte en langage naturel, entra\u00een\u00e9 sur un vaste volume de documents).<\/span><\/p><p><span style=\"font-weight: 400;\">En tant que LLM, <a href=\"https:\/\/datascientest.com\/meta-se-bat-a-armes-egales-contre-openai-et-gpt-4\">LLaMa est comparable \u00e0 GPT<\/a>, le mod\u00e8le sur lequel s\u2019appuie <a href=\"https:\/\/datascientest.com\/comment-fonctionne-chatgpt\">ChatGPT<\/a> ou encore <a href=\"https:\/\/datascientest.com\/google-deepmind-compte-detroner-chatgpt-mais-pas-avec-bard\">DeepMind<\/a>, celui sur lequel s\u2019appuie <a href=\"https:\/\/datascientest.com\/google-gemini-tout-savoir\">Google Gemini<\/a>. Pourtant, il se veut diff\u00e9rent\u2026<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7634ada elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"7634ada\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-intelligence-artificielle\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Savoir d\u00e9velopper une intelligence artificielle<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-97aad59 e-flex e-con-boxed e-con\" data-id=\"97aad59\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-09faa3c elementor-widget elementor-widget-heading\" data-id=\"09faa3c\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">LLaMa, le crois\u00e9 de l\u2019open source<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2d2ea89 elementor-widget elementor-widget-text-editor\" data-id=\"2d2ea89\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Ce qui caract\u00e9rise LLaMa est qu\u2019il s\u2019agit d\u2019<strong>un LLM open source<\/strong>. Cela signifie qu\u2019il est accessible gratuitement, mais aussi que n\u2019importe qui peut intervenir sur le code et affiner certains aspects.\u00a0 Un projet open source peut <strong>b\u00e9n\u00e9ficier des apports de milliers de d\u00e9veloppeurs<\/strong> hors pair. C\u2019est un des aspects qui a fait la popularit\u00e9 de Linux : la s\u00e9curit\u00e9 a \u00e9t\u00e9 renforc\u00e9e par des arm\u00e9es de contributeurs \u00e0 l\u2019aff\u00fbt.<\/span><\/p><p><span style=\"font-weight: 400;\">Au niveau de l\u2019utilisateur, un avantage vient de ce qu\u2019il est possible de le t\u00e9l\u00e9charger int\u00e9gralement sur son ordinateur \u2013 sachant qu\u2019il faut tout de m\u00eame une sacr\u00e9e configuration pour la version la plus avanc\u00e9e \u2013 et donc l\u2019utiliser sans connexion Internet. Il faut juste savoir que ses connaissances sont limit\u00e9es \u00e0 d\u00e9cembre 2023.<\/span><\/p><p><span style=\"font-weight: 400;\">Pourtant, LLaMa est un LLM produit par <a href=\"https:\/\/datascientest.com\/facebook-devient-meta-et-va-recruter-10-000-personnes-en-europe-pour-construire-le-metaverse\">la soci\u00e9t\u00e9 Meta<\/a>, et donc celle qui g\u00e8re <strong>Facebook<\/strong>, <strong>Instagram<\/strong> ou <strong>Whatsapp<\/strong>. Et des centaines de millions de dollars ont \u00e9t\u00e9 investis dans son d\u00e9veloppement.<\/span><\/p><p><span style=\"font-weight: 400;\">Pourquoi Zuckerberg a-t-il choisi une telle approche ? Nous y reviendrons.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-a26e509 elementor-widget elementor-widget-image\" data-id=\"a26e509\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-image{text-align:center}.elementor-widget-image a{display:inline-block}.elementor-widget-image a img[src$=\".svg\"]{width:48px}.elementor-widget-image img{vertical-align:middle;display:inline-block}<\/style>\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1329\" height=\"666\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_1.jpg\" class=\"attachment-full size-full wp-image-251537\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_1.jpg 1329w, https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_1-300x150.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_1-1024x513.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_1-768x385.jpg 768w\" sizes=\"(max-width: 1329px) 100vw, 1329px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-1446644 e-flex e-con-boxed e-con\" data-id=\"1446644\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-f0b78f5 elementor-widget elementor-widget-heading\" data-id=\"f0b78f5\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Des performances acclam\u00e9es<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1d21757 elementor-widget elementor-widget-text-editor\" data-id=\"1d21757\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Le facteur open source serait n\u00e9gligeable si les performances n\u2019\u00e9taient pas au rendez-vous. Or, \u00e0 chaque version, LLaMa a d\u00e9multipli\u00e9 ses capacit\u00e9s. Il est vrai aussi que les versions se sont succ\u00e9d\u00e9 rapidement :<\/span><\/p><ul><li><span style=\"font-weight: 400;\"><strong>LLaMa 1<\/strong> : f\u00e9vrier 2023<\/span><\/li><li><span style=\"font-weight: 400;\"><a href=\"https:\/\/datascientest.com\/meta-devoile-llama-2-mais-cette-fois-en-open-source-pourquoi-est-ce-important\"><strong>LLaMa 2<\/strong><\/a> : juillet 2023<\/span><\/li><li><span style=\"font-weight: 400;\"><a href=\"https:\/\/datascientest.com\/meta-presente-llama-3-le-llm-gratuit-qui-repousse-les-limites-du-marche\"><strong>LLaMa 3<\/strong><\/a> : avril 2023 &#8211; disponible en deux versions : 8B (8 milliards de param\u00e8tres) et 70B (70 milliards).<\/span><\/li><li><span style=\"font-weight: 400;\"><strong>LLaMa 3.1<\/strong> : fin juillet 2024<\/span><\/li><\/ul>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-230366b elementor-widget elementor-widget-text-editor\" data-id=\"230366b\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">C\u2019est \u00e0 partir de la version 3 que LLaMa a commenc\u00e9 \u00e0 faire l&rsquo;objet d\u2019\u00e9loges appuy\u00e9s, se voyant qualifi\u00e9 par certains de <strong>\u00ab surpuissant \u00bb<\/strong>. Et ses performances ont \u00e9t\u00e9 reconnues par des benchmark comme MMLu. Les taux de r\u00e9ussite de 3 des principaux LLM \u00e9taient les suivants :<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-8bf6fd3 elementor-widget elementor-widget-html\" data-id=\"8bf6fd3\" data-element_type=\"widget\" data-widget_type=\"html.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style type=\"text\/css\">\r\n.tg  {border-collapse:collapse;border-spacing:0;}\r\n.tg td{border-color:black;border-style:solid;border-width:1px;font-family:Rubik, sans-serif;font-size:14px;\r\n  overflow:hidden;padding:10px 5px;word-break:normal;}\r\n.tg th{border-color:black;border-style:solid;border-width:1px;font-family:Rubik, sans-serif;font-size:14px;\r\n  font-weight:normal;overflow:hidden;padding:10px 5px;word-break:normal;}\r\n.tg .tg-na62{font-size:18px;font-weight:bold;text-align:left;vertical-align:middle}\r\n.tg .tg-px4u{font-size:18px;text-align:right;vertical-align:middle}\r\n<\/style>\r\n<table class=\"tg\" style=\"undefined;table-layout: fixed; width: 250px\"><colgroup>\r\n<col style=\"width: 150px\">\r\n<col style=\"width: 100px\">\r\n<\/colgroup>\r\n<thead>\r\n  <tr>\r\n    <th class=\"tg-na62\"><span style=\"font-style:normal;text-decoration:none;color:#000;background-color:transparent\">Gemini Ultra<\/span><\/th>\r\n    <th class=\"tg-px4u\"><span style=\"font-weight:400;font-style:normal;text-decoration:none;color:#000;background-color:transparent\">90 %<\/span><\/th>\r\n  <\/tr><\/thead>\r\n<tbody>\r\n  <tr>\r\n    <td class=\"tg-na62\"><span style=\"font-style:normal;text-decoration:none;color:#000;background-color:transparent\">GPT-4<\/span><\/td>\r\n    <td class=\"tg-px4u\"><span style=\"font-weight:400;font-style:normal;text-decoration:none;color:#000;background-color:transparent\">86 %<\/span><\/td>\r\n  <\/tr>\r\n  <tr>\r\n    <td class=\"tg-na62\"><span style=\"font-style:normal;text-decoration:none;color:#000;background-color:transparent\">LLaMa 70B<\/span><\/td>\r\n    <td class=\"tg-px4u\"><span style=\"font-weight:400;font-style:normal;text-decoration:none;color:#000;background-color:transparent\">82 %<\/span><\/td>\r\n  <\/tr>\r\n<\/tbody>\r\n<\/table>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-cf90f78 elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"cf90f78\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-intelligence-artificielle\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Se former \u00e0 l'intelligence artificielle<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-032b396 elementor-widget elementor-widget-text-editor\" data-id=\"032b396\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">LLaMa 3.1 a repr\u00e9sent\u00e9 un pas en avant consid\u00e9rable, puisque <strong>la version 405 B<\/strong> repose sur <strong>405 milliards de param\u00e8tres<\/strong>. Cette fois, au test MMLu, le score obtenu (88,6 %) est \u00e0 peine inf\u00e9rieur \u00e0 celui de <a href=\"https:\/\/datascientest.com\/gpt-4o-openai-donne-de-la-voix-a-lia-avec-son-nouveau-modele\">GTP4-o<\/a> (88,7 %).<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-407d6d8 e-flex e-con-boxed e-con\" data-id=\"407d6d8\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-88ce136 elementor-widget elementor-widget-heading\" data-id=\"88ce136\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Pourquoi Meta a-t-elle choisi la formule open source ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-17cd6c8 elementor-widget elementor-widget-text-editor\" data-id=\"17cd6c8\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Mark Zuckerberg a profit\u00e9 de LLaMa pour se faire conna\u00eetre comme un supporter de la formule open source. En r\u00e9alit\u00e9, Meta se retrouve gagnant \u00e0 plus d\u2019un titre.<\/span><\/p><ul><li><span style=\"font-weight: 400;\">Dans la <strong>Silicon Valley<\/strong>, les pr\u00e9ceptes de Sun Tzu sont appr\u00e9ci\u00e9s et l\u2019un d\u2019eux est le suivant : \u00ab si tu ne peux pas vaincre sur un champ de bataille, change le champ de bataille. \u00bb Or, <strong>en proposant LLaMa gratuitement<\/strong>, Meta coupe l\u2019herbe sous le pied de ses trois principaux concurrents : <a href=\"https:\/\/datascientest.com\/openai-lentreprise-qui-revolutionne-lia\">OpenAI<\/a>, <strong>Google<\/strong> et <a href=\"https:\/\/datascientest.com\/anthropic-developpe-un-nouveau-systeme-dia-le-golden-gate-bridge\">Anthropic<\/a> (<a href=\"https:\/\/datascientest.com\/claude-3-danthropic-le-llm-qui-detrone-gpt-4\">Claude<\/a>). Et ce, avant qu\u2019ils n\u2019aient plus rafler l\u2019essentiel du march\u00e9 de l\u2019IA.<\/span><\/li><li><span style=\"font-weight: 400;\">Au passage, Meta peut redorer son blason, suite aux affaires de n\u00e9gligence de la vie priv\u00e9e qui ont entach\u00e9 son parcours.<\/span><\/li><li><span style=\"font-weight: 400;\">De par la nature de l\u2019open source, LLaMa peut b\u00e9n\u00e9ficier des <strong>retours de milliers de d\u00e9veloppeurs.<\/strong><\/span><\/li><\/ul>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6a8439f elementor-widget elementor-widget-image\" data-id=\"6a8439f\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1310\" height=\"669\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_2.jpg\" class=\"attachment-full size-full wp-image-251538\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_2.jpg 1310w, https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_2-300x153.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_2-1024x523.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest_2-768x392.jpg 768w\" sizes=\"(max-width: 1310px) 100vw, 1310px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-fe400be elementor-widget elementor-widget-text-editor\" data-id=\"fe400be\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<ul><li><span style=\"font-weight: 400;\">Le c\u00f4t\u00e9 <strong>open source<\/strong> et donc non propri\u00e9taire peut rassurer bien des soci\u00e9t\u00e9s inqui\u00e8tes de ce que des IA comme ChatGPT ou Gemini puissent analyser leurs informations. Ainsi, Samsung, Amazon ou Apple ont fait interdire l\u2019usage de ChatGPT d\u00e8s la fin du printemps 2023. Il en a \u00e9t\u00e9 de m\u00eame pour divers cabinets d\u2019avocats, h\u00f4pitaux ou organisations gouvernementales.<\/span><\/li><li><span style=\"font-weight: 400;\">Meta est en mesure de proposer <strong>l\u2019int\u00e9gration de LLaMa 3<\/strong> \u00e0 des entreprises dans le cadre de t\u00e2ches sur mesure, tout en se pr\u00e9valant d\u2019une technologie \u00e9prouv\u00e9e.<\/span><\/li><li><span style=\"font-weight: 400;\">L\u00e0 o\u00f9 <a href=\"https:\/\/datascientest.com\/differences-clouds\">Amazon, Microsoft et Google<\/a> se sont impos\u00e9s dans le cloud, Meta trouve dans l\u2019IA un domaine o\u00f9 elle peut s\u2019imposer \u00e0 grande \u00e9chelle face aux autres <strong>GAFAM<\/strong>.<\/span><\/li><li><span style=\"font-weight: 400;\">Zuckerberg se voit d\u00e9sormais interview\u00e9 en tant que guru de l\u2019open source et a beau jeu de se livrer \u00e0 des d\u00e9clarations telles que \u00ab l\u2019IA open source va devenir le standard de l\u2019industrie \u00bb. \u00ab <strong>LLaMa n\u2019est pas une technologie mais un \u00e9cosyst\u00e8me<\/strong> \u00bb\u2026 Et de fait, LLaMa 3 a rapidement \u00e9t\u00e9 int\u00e9gr\u00e9 dans l\u2019offre de g\u00e9ants tels que Amazon Web Services et Microsoft Azure, Zoom ou encore AT&amp;T.<\/span><\/li><\/ul>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-6b103bf e-flex e-con-boxed e-con\" data-id=\"6b103bf\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-2fe1b44 elementor-widget elementor-widget-heading\" data-id=\"2fe1b44\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">En fran\u00e7ais dans le texte ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6e16d2c elementor-widget elementor-widget-text-editor\" data-id=\"6e16d2c\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Pour toucher le grand public, il faut une interface aussi simple d\u2019usage que celle de ChatGPT. Elle existe sous la forme d\u2019une app, nomm\u00e9e <strong>Meta AI<\/strong>. Tout comme ChatGPT, elle peut g\u00e9n\u00e9rer des images et m\u00eame les animer.<\/span><\/p><p><span style=\"font-weight: 400;\">Toutefois, ce qui a pu freiner l\u2019audience de Meta AI est que, d\u00e9but septembre 2024, <strong>cette app n\u2019\u00e9tait pas encore accessible en Europe<\/strong> (officiellement, en raison de probl\u00e8mes de r\u00e9gulations) et donc en France. <\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-0784b1c elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"0784b1c\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-intelligence-artificielle\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Tout apprendre de l'intelligence artificielle<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-75aee8a e-flex e-con-boxed e-con\" data-id=\"75aee8a\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-4b05e7a elementor-widget elementor-widget-html\" data-id=\"4b05e7a\" data-element_type=\"widget\" data-widget_type=\"html.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<script type=\"application\/ld+json\">\r\n{\r\n  \"@context\": \"https:\/\/schema.org\",\r\n  \"@type\": \"FAQPage\",\r\n  \"mainEntity\": [{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Qu'est-ce que LLaMa ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"LLaMa (Large Language Model Meta AI) est une s\u00e9rie de mod\u00e8les de langage d\u00e9velopp\u00e9s par Meta (anciennement Facebook). Ces mod\u00e8les sont con\u00e7us pour comprendre et g\u00e9n\u00e9rer du texte en langage naturel, permettant des applications vari\u00e9es comme la g\u00e9n\u00e9ration de texte, la traduction automatique, et la r\u00e9ponse \u00e0 des questions.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Quels sont les avantages des mod\u00e8les LLaMa ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Les mod\u00e8les LLaMa se distinguent par leur efficacit\u00e9 et leur capacit\u00e9 \u00e0 comprendre des contextes complexes. Ils sont optimis\u00e9s pour des t\u00e2ches de traitement du langage naturel avec une architecture moderne qui permet des performances \u00e9lev\u00e9es tout en restant relativement l\u00e9gers en termes de ressources n\u00e9cessaires.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Comment sont entra\u00een\u00e9s les mod\u00e8les LLaMa ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Les mod\u00e8les LLaMa sont entra\u00een\u00e9s en utilisant des ensembles de donn\u00e9es divers et volumineux pour apprendre les structures linguistiques, les faits, et les contextes. L\u2019entra\u00eenement implique l\u2019utilisation de techniques avanc\u00e9es d\u2019apprentissage supervis\u00e9 et non supervis\u00e9 pour am\u00e9liorer la capacit\u00e9 du mod\u00e8le \u00e0 g\u00e9n\u00e9rer du texte pertinent et coh\u00e9rent.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"O\u00f9 peut-on utiliser LLaMa ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Les mod\u00e8les LLaMa peuvent \u00eatre int\u00e9gr\u00e9s dans diverses applications telles que les chatbots, les assistants virtuels, les outils de g\u00e9n\u00e9ration de contenu, et les syst\u00e8mes de recommandation. Ils sont \u00e9galement utilis\u00e9s dans la recherche et le d\u00e9veloppement de nouvelles technologies en intelligence artificielle pour am\u00e9liorer la compr\u00e9hension du langage et la g\u00e9n\u00e9ration de texte.\"\r\n    }\r\n  }]\r\n}\r\n<\/script>\r\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>Contre toute attente, LLaMa, le mod\u00e8le d\u2019intelligence artificielle propos\u00e9 par Meta, conna\u00eet un succ\u00e8s remarquable. Au risque de faire de l\u2019ombre \u00e0 ChatGPT, Google Gemini et autres stars du lot. Comment expliquer cette mont\u00e9e en popularit\u00e9 rapide et soudaine ? Pour un succ\u00e8s, c\u2019est un succ\u00e8s. Fin ao\u00fbt 2024, LLaMa d\u00e9nombrait 350 millions de t\u00e9l\u00e9chargements, [&hellip;]<\/p>\n","protected":false},"author":77,"featured_media":251521,"comment_status":"open","ping_status":"open","sticky":false,"template":"elementor_theme","format":"standard","meta":{"content-type":"","inline_featured_image":false},"categories":[2500],"tags":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v21.7 (Yoast SEO v21.7) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>LLaMa : tout savoir sur le mod\u00e8le de langage de Meta<\/title>\n<meta name=\"description\" content=\"D\u00e9couvrez LLaMa, un mod\u00e8le de langage avanc\u00e9 d\u00e9velopp\u00e9 par Meta, qui red\u00e9finit les capacit\u00e9s des intelligences artificielles \u00e0 grande \u00e9chelle. Explorez son architecture, ses fonctionnalit\u00e9s innovantes, et son impact sur le traitement du langage naturel et les applications IA.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/datascientest.com\/llama-tout-savoir\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"LLaMa : tout savoir sur le mod\u00e8le de langage de Meta\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/llama-tout-savoir\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2024-09-25T05:30:00+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-10-03T11:01:54+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"600\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Raphael Kassel\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"Raphael Kassel\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"5 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/llama-tout-savoir\",\"url\":\"https:\/\/datascientest.com\/llama-tout-savoir\",\"name\":\"LLaMa : tout savoir sur le mod\u00e8le de langage de Meta\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2024-09-25T05:30:00+00:00\",\"dateModified\":\"2024-10-03T11:01:54+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1\"},\"description\":\"D\u00e9couvrez LLaMa, un mod\u00e8le de langage avanc\u00e9 d\u00e9velopp\u00e9 par Meta, qui red\u00e9finit les capacit\u00e9s des intelligences artificielles \u00e0 grande \u00e9chelle. Explorez son architecture, ses fonctionnalit\u00e9s innovantes, et son impact sur le traitement du langage naturel et les applications IA.\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/llama-tout-savoir\"]}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/datascientest.com\/#website\",\"url\":\"https:\/\/datascientest.com\/\",\"name\":\"Formation Data Science | DataScientest.com\",\"description\":\"Leader de la formation en Data Science\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/datascientest.com\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1\",\"name\":\"Raphael Kassel\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg\",\"caption\":\"Raphael Kassel\"},\"description\":\"Raphael Kassel est le Head of Data Programs chez DataScientest. Il supervise une \u00e9quipe Data, impl\u00e9mente de nouveaux parcours comme \\\"Chef de projet IA\\\" et \\\"Data Product Manager\\\", et contribue \u00e0 la modernisation des op\u00e9rations du groupe. En collaboration \u00e9troite avec l'\u00e9quipe de d\u00e9veloppement, il adapte l'accompagnement acad\u00e9mique aux \u00e9volutions de l'\u00e9quipe data et \u00e0 l'internationalisation.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/raphaelkassel\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/raphaelkassel\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"LLaMa : tout savoir sur le mod\u00e8le de langage de Meta","description":"D\u00e9couvrez LLaMa, un mod\u00e8le de langage avanc\u00e9 d\u00e9velopp\u00e9 par Meta, qui red\u00e9finit les capacit\u00e9s des intelligences artificielles \u00e0 grande \u00e9chelle. Explorez son architecture, ses fonctionnalit\u00e9s innovantes, et son impact sur le traitement du langage naturel et les applications IA.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/datascientest.com\/llama-tout-savoir","og_locale":"fr_FR","og_type":"article","og_title":"LLaMa : tout savoir sur le mod\u00e8le de langage de Meta","og_url":"https:\/\/datascientest.com\/llama-tout-savoir","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2024-09-25T05:30:00+00:00","article_modified_time":"2024-10-03T11:01:54+00:00","og_image":[{"width":1200,"height":600,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2024\/09\/llama_datascientest.png","type":"image\/png"}],"author":"Raphael Kassel","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"Raphael Kassel","Dur\u00e9e de lecture estim\u00e9e":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/llama-tout-savoir","url":"https:\/\/datascientest.com\/llama-tout-savoir","name":"LLaMa : tout savoir sur le mod\u00e8le de langage de Meta","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2024-09-25T05:30:00+00:00","dateModified":"2024-10-03T11:01:54+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1"},"description":"D\u00e9couvrez LLaMa, un mod\u00e8le de langage avanc\u00e9 d\u00e9velopp\u00e9 par Meta, qui red\u00e9finit les capacit\u00e9s des intelligences artificielles \u00e0 grande \u00e9chelle. Explorez son architecture, ses fonctionnalit\u00e9s innovantes, et son impact sur le traitement du langage naturel et les applications IA.","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/llama-tout-savoir"]}]},{"@type":"WebSite","@id":"https:\/\/datascientest.com\/#website","url":"https:\/\/datascientest.com\/","name":"Formation Data Science | DataScientest.com","description":"Leader de la formation en Data Science","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/datascientest.com\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1","name":"Raphael Kassel","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg","caption":"Raphael Kassel"},"description":"Raphael Kassel est le Head of Data Programs chez DataScientest. Il supervise une \u00e9quipe Data, impl\u00e9mente de nouveaux parcours comme \"Chef de projet IA\" et \"Data Product Manager\", et contribue \u00e0 la modernisation des op\u00e9rations du groupe. En collaboration \u00e9troite avec l'\u00e9quipe de d\u00e9veloppement, il adapte l'accompagnement acad\u00e9mique aux \u00e9volutions de l'\u00e9quipe data et \u00e0 l'internationalisation.","sameAs":["https:\/\/www.linkedin.com\/in\/raphaelkassel\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/raphaelkassel"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/251496"}],"collection":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/users\/77"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=251496"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/251496\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/251521"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=251496"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=251496"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=251496"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}