{"id":2258,"date":"2026-04-16T10:27:46","date_gmt":"2026-04-16T08:27:46","guid":{"rendered":"https:\/\/askem.eu\/?p=2258"},"modified":"2026-04-16T10:27:52","modified_gmt":"2026-04-16T08:27:52","slug":"unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste","status":"publish","type":"post","link":"https:\/\/askem.eu\/en\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/","title":{"rendered":"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste"},"content":{"rendered":"<h1 class=\"wp-block-heading\">Unsloth&nbsp;: fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste<\/h1>\n\n\n\n<p>Comment adapter un mod\u00e8le de langage \u00e0 ses donn\u00e9es m\u00e9tier en quelques heures, avec LoRA et QLoRA, sans cluster GPU et l&rsquo;int\u00e9grer dans sa stack open source existante.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Pourquoi fine-tuner un LLM&nbsp;?<\/h2>\n\n\n\n<p>Ollama et vLLM permettent de servir des mod\u00e8les open source comme Llama 3, Mistral ou Qwen en production. Mais un mod\u00e8le g\u00e9n\u00e9raliste ne conna\u00eet pas votre terminologie m\u00e9tier, vos formats de sortie, ni le ton attendu par vos utilisateurs. Le fine-tuning (ajustement fin) consiste \u00e0 r\u00e9-entra\u00eener partiellement un mod\u00e8le sur vos propres donn\u00e9es pour qu&rsquo;il devienne sp\u00e9cialiste de votre domaine, sans repartir de z\u00e9ro.<\/p>\n\n\n\n<p>Les cas d&rsquo;usage concrets sont nombreux&nbsp;: un LLM qui r\u00e9pond aux questions sur la documentation interne d&rsquo;une collectivit\u00e9, un mod\u00e8le qui g\u00e9n\u00e8re des fiches DCAT conformes \u00e0 votre sch\u00e9ma de m\u00e9tadonn\u00e9es, un assistant qui produit du JSON structur\u00e9 selon votre API, ou encore un chatbot qui adopte le vocabulaire d&rsquo;un secteur technique pr\u00e9cis.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Unsloth&nbsp;: le fine-tuning accessible<\/h2>\n\n\n\n<p><a href=\"https:\/\/unsloth.ai\" type=\"link\" id=\"https:\/\/unsloth.ai\">Unsloth<\/a> est une biblioth\u00e8que Python open source (licence Apache 2.0) qui acc\u00e9l\u00e8re le fine-tuning de LLM d&rsquo;un facteur 2 \u00e0 5x tout en r\u00e9duisant la consommation m\u00e9moire GPU de 70 \u00e0 80&nbsp;%. Concr\u00e8tement, un fine-tuning qui n\u00e9cessitait un GPU A100 80 Go devient r\u00e9alisable sur un GPU grand public de 16 \u00e0 24 Go (RTX 4090, RTX 3090, ou m\u00eame un T4 gratuit sur Google Colab).<\/p>\n\n\n\n<p>Le projet repose sur des optimisations noyau (kernels Triton compil\u00e9s manuellement) qui acc\u00e9l\u00e8rent les op\u00e9rations critiques de l&rsquo;entra\u00eenement&nbsp;: attention, normalisation RMS, cross-entropy, sans compromettre la pr\u00e9cision num\u00e9rique. Unsloth ne modifie pas l&rsquo;architecture du mod\u00e8le&nbsp;: il rend le processus d&rsquo;entra\u00eenement plus efficace.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">LoRA et QLoRA&nbsp;: adapter sans tout r\u00e9\u00e9crire<\/h2>\n\n\n\n<p>Le fine-tuning complet d&rsquo;un LLM de 7 milliards de param\u00e8tres exige des centaines de gigaoctets de m\u00e9moire. LoRA (Low-Rank Adaptation) contourne cette contrainte en gelant les poids originaux du mod\u00e8le et en entra\u00eenant uniquement de petites matrices de rang faible inject\u00e9es dans les couches d&rsquo;attention. On n&rsquo;entra\u00eene ainsi que 0,1 \u00e0 1&nbsp;% des param\u00e8tres, pour un r\u00e9sultat souvent comparable au fine-tuning complet.<\/p>\n\n\n\n<p>QLoRA pousse l&rsquo;optimisation plus loin en quantifiant le mod\u00e8le de base en 4 bits (NF4) avant d&rsquo;appliquer LoRA. Le mod\u00e8le gel\u00e9 occupe quatre fois moins de m\u00e9moire, et seuls les adaptateurs LoRA restent en pr\u00e9cision compl\u00e8te. C&rsquo;est cette combinaison&nbsp;: quantification 4 bits du mod\u00e8le de base + adaptateurs LoRA en bf16 qu&rsquo;Unsloth optimise et simplifie.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Pr\u00e9parer un jeu de donn\u00e9es d&rsquo;entra\u00eenement<\/h2>\n\n\n\n<p>La qualit\u00e9 du fine-tuning d\u00e9pend directement de la qualit\u00e9 des donn\u00e9es. Unsloth attend un jeu de donn\u00e9es au format conversationnel (liste de messages avec r\u00f4les system\/user\/assistant) ou au format instruction\/output. Quelques centaines d&rsquo;exemples de haute qualit\u00e9 suffisent pour un fine-tuning efficace par LoRA, inutile de viser des millions de lignes.<\/p>\n\n\n\n<p>Pour constituer ce jeu de donn\u00e9es, plusieurs approches sont possibles&nbsp;: annoter manuellement des exemples repr\u00e9sentatifs, utiliser un LLM plus puissant pour g\u00e9n\u00e9rer des paires question-r\u00e9ponse \u00e0 partir de documents existants (distillation synth\u00e9tique), ou extraire et reformater des donn\u00e9es existantes avec Docling. Le jeu de donn\u00e9es peut \u00eatre stock\u00e9 en JSON, CSV ou directement au format Hugging Face Datasets.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Workflow type avec Unsloth<\/h2>\n\n\n\n<p>Un fine-tuning typique avec Unsloth se d\u00e9roule en cinq \u00e9tapes&nbsp;:<\/p>\n\n\n\n<p><strong>1. Charger le mod\u00e8le de base quantifi\u00e9.<\/strong> Unsloth fournit des versions pr\u00e9-quantifi\u00e9es des mod\u00e8les populaires (Llama 3.1, Mistral v0.3, Qwen 2.5, Phi-3, Gemma 2) pr\u00eates \u00e0 l&#8217;emploi. Un appel \u00e0 <code>FastLanguageModel.from_pretrained()<\/code> charge le mod\u00e8le en 4 bits avec les optimisations Unsloth activ\u00e9es.<\/p>\n\n\n\n<p><strong>2. Appliquer les adaptateurs LoRA.<\/strong> On sp\u00e9cifie le rang (r=16 ou r=32 typiquement), les couches cibles (q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj), et le facteur alpha. Unsloth optimise automatiquement les gradients pour ces couches.<\/p>\n\n\n\n<p><strong>3. Formater le jeu de donn\u00e9es.<\/strong> Les donn\u00e9es sont converties au format attendu par le mod\u00e8le via un template de chat (ChatML, Llama-3, Mistral). Unsloth inclut des helpers pour les formats courants.<\/p>\n\n\n\n<p><strong>4. Lancer l&rsquo;entra\u00eenement.<\/strong> Unsloth s&rsquo;appuie sur le Trainer de Hugging Face (SFTTrainer de la biblioth\u00e8que TRL). On configure les hyperparam\u00e8tres classiques&nbsp;: taux d&rsquo;apprentissage (2e-4 typiquement), nombre d&rsquo;\u00e9poques (1 \u00e0 3), taille de batch, warmup. L&rsquo;entra\u00eenement sur 1 000 exemples avec un mod\u00e8le 7B prend environ 20 \u00e0 40 minutes sur un GPU 24 Go.<\/p>\n\n\n\n<p><strong>5. Exporter et d\u00e9ployer.<\/strong> Le mod\u00e8le fine-tun\u00e9 peut \u00eatre export\u00e9 en plusieurs formats&nbsp;: adaptateurs LoRA seuls (quelques Mo), mod\u00e8le fusionn\u00e9 en fp16, ou directement en GGUF pour Ollama. Cette derni\u00e8re option permet de d\u00e9ployer imm\u00e9diatement le mod\u00e8le personnalis\u00e9 dans la stack existante avec <code>ollama create<\/code>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Int\u00e9gration dans la stack existante<\/h2>\n\n\n\n<p>Un mod\u00e8le fine-tun\u00e9 avec Unsloth s&rsquo;int\u00e8gre naturellement dans l&rsquo;\u00e9cosyst\u00e8me d\u00e9j\u00e0 couvert sur ce site. L&rsquo;export GGUF se charge directement dans <a href=\"https:\/\/askem.eu\/en\/2026\/03\/29\/ollama-executer-des-llm-en-local\/\" type=\"post\" id=\"2141\">Ollama<\/a> ou <a href=\"https:\/\/askem.eu\/en\/2026\/04\/07\/vllm-servir-des-llm-a-haute-performance-en-production\/\" type=\"post\" id=\"2198\">vLLM<\/a> pour le serving. <a href=\"https:\/\/askem.eu\/en\/2026\/04\/08\/litellm-un-proxy-unifie-pour-router-ses-requetes-llm-entre-ollama-vllm-et-le-cloud\/\" type=\"post\" id=\"2206\">LiteLLM<\/a> peut router certaines requ\u00eates vers le mod\u00e8le sp\u00e9cialis\u00e9 et d&rsquo;autres vers un mod\u00e8le g\u00e9n\u00e9raliste. <a href=\"https:\/\/askem.eu\/en\/2026\/04\/02\/langfuse-observer-et-evaluer-ses-pipelines-llm-open-source-en-production\/\" type=\"post\" id=\"2162\">Langfuse<\/a> permet de comparer les performances du mod\u00e8le fine-tun\u00e9 par rapport au mod\u00e8le de base sur des m\u00e9triques concr\u00e8tes (qualit\u00e9 des r\u00e9ponses, respect du format, pertinence m\u00e9tier). Et <a href=\"https:\/\/askem.eu\/en\/2026\/03\/30\/n8n-automatiser-ses-workflows\/\" type=\"post\" id=\"2144\">n8n<\/a> ou <a href=\"https:\/\/askem.eu\/en\/2026\/04\/09\/langgraph-construire-des-agents-ia-autonomes-avec-orchestration-memoire-et-outils\/\" type=\"post\" id=\"2212\">LangGraph<\/a> orchestrent les workflows qui appellent le mod\u00e8le adapt\u00e9.<\/p>\n\n\n\n<p>Pour \u00e9valuer objectivement le gain, une bonne pratique consiste \u00e0 constituer un jeu de test s\u00e9par\u00e9 (50 \u00e0 100 exemples annot\u00e9s) et \u00e0 comparer les r\u00e9ponses du mod\u00e8le de base et du mod\u00e8le fine-tun\u00e9 avec Langfuse. Les m\u00e9triques \u00e0 suivre d\u00e9pendent du cas d&rsquo;usage&nbsp;: respect du format JSON (mesurable avec <a href=\"https:\/\/askem.eu\/en\/2026\/04\/13\/outlines-forcer-les-llm-open-source-a-produire-du-json-structure-en-production\/\" type=\"post\" id=\"2242\">Outlines<\/a>), pertinence des r\u00e9ponses RAG (\u00e9valuable avec RAGAS), ou score humain sur un \u00e9chantillon.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Limites et bonnes pratiques<\/h2>\n\n\n\n<p>Le fine-tuning n&rsquo;est pas une solution universelle. Il est inutile si le prompting ou le RAG suffisent \u00e0 obtenir le r\u00e9sultat souhait\u00e9&nbsp;: toujours tester ces approches d&rsquo;abord. Il est contre-productif sur un jeu de donn\u00e9es trop petit (moins de 50 exemples) ou de mauvaise qualit\u00e9 (r\u00e9ponses incoh\u00e9rentes, erreurs factuelles). Et il introduit un risque de surapprentissage (overfitting)&nbsp;: le mod\u00e8le peut m\u00e9moriser les exemples au lieu de g\u00e9n\u00e9raliser.<\/p>\n\n\n\n<p>Quelques garde-fous essentiels&nbsp;: toujours s\u00e9parer donn\u00e9es d&rsquo;entra\u00eenement et donn\u00e9es de test, surveiller la loss de validation pendant l&rsquo;entra\u00eenement, limiter le nombre d&rsquo;\u00e9poques (1 \u00e0 3 suffisent g\u00e9n\u00e9ralement), et versionner les adaptateurs LoRA comme du code (ils ne font que quelques Mo et se stockent facilement dans Git).<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Ressources<\/h2>\n\n\n\n<p>Le d\u00e9p\u00f4t GitHub <a href=\"https:\/\/github.com\/unslothai\/unsloth\">d&rsquo;Unsloth<\/a> contient des notebooks Colab pr\u00eats \u00e0 l&#8217;emploi pour chaque famille de mod\u00e8les, permettant de tester le fine-tuning en quelques minutes sans infrastructure locale. La documentation couvre l&rsquo;ensemble du workflow, de la pr\u00e9paration des donn\u00e9es \u00e0 l&rsquo;export GGUF. Le wiki du projet d\u00e9taille les hyperparam\u00e8tres recommand\u00e9s par taille de mod\u00e8le et par type de t\u00e2che.<\/p>\n\n\n\n<p><\/p>","protected":false},"excerpt":{"rendered":"<p>Unsloth&nbsp;: fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste Comment adapter un mod\u00e8le de langage \u00e0 ses donn\u00e9es m\u00e9tier en quelques heures, avec LoRA et QLoRA, sans cluster GPU et l&rsquo;int\u00e9grer dans sa stack open source existante. Pourquoi fine-tuner un LLM&nbsp;? Ollama et vLLM permettent de servir des mod\u00e8les open source comme Llama [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":2259,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ocean_post_layout":"","ocean_both_sidebars_style":"","ocean_both_sidebars_content_width":0,"ocean_both_sidebars_sidebars_width":0,"ocean_sidebar":"","ocean_second_sidebar":"","ocean_disable_margins":"enable","ocean_add_body_class":"","ocean_shortcode_before_top_bar":"","ocean_shortcode_after_top_bar":"","ocean_shortcode_before_header":"","ocean_shortcode_after_header":"","ocean_has_shortcode":"","ocean_shortcode_after_title":"","ocean_shortcode_before_footer_widgets":"","ocean_shortcode_after_footer_widgets":"","ocean_shortcode_before_footer_bottom":"","ocean_shortcode_after_footer_bottom":"","ocean_display_top_bar":"default","ocean_display_header":"default","ocean_header_style":"","ocean_center_header_left_menu":"","ocean_custom_header_template":"","ocean_custom_logo":0,"ocean_custom_retina_logo":0,"ocean_custom_logo_max_width":0,"ocean_custom_logo_tablet_max_width":0,"ocean_custom_logo_mobile_max_width":0,"ocean_custom_logo_max_height":0,"ocean_custom_logo_tablet_max_height":0,"ocean_custom_logo_mobile_max_height":0,"ocean_header_custom_menu":"","ocean_menu_typo_font_family":"","ocean_menu_typo_font_subset":"","ocean_menu_typo_font_size":0,"ocean_menu_typo_font_size_tablet":0,"ocean_menu_typo_font_size_mobile":0,"ocean_menu_typo_font_size_unit":"px","ocean_menu_typo_font_weight":"","ocean_menu_typo_font_weight_tablet":"","ocean_menu_typo_font_weight_mobile":"","ocean_menu_typo_transform":"","ocean_menu_typo_transform_tablet":"","ocean_menu_typo_transform_mobile":"","ocean_menu_typo_line_height":0,"ocean_menu_typo_line_height_tablet":0,"ocean_menu_typo_line_height_mobile":0,"ocean_menu_typo_line_height_unit":"","ocean_menu_typo_spacing":0,"ocean_menu_typo_spacing_tablet":0,"ocean_menu_typo_spacing_mobile":0,"ocean_menu_typo_spacing_unit":"","ocean_menu_link_color":"","ocean_menu_link_color_hover":"","ocean_menu_link_color_active":"","ocean_menu_link_background":"","ocean_menu_link_hover_background":"","ocean_menu_link_active_background":"","ocean_menu_social_links_bg":"","ocean_menu_social_hover_links_bg":"","ocean_menu_social_links_color":"","ocean_menu_social_hover_links_color":"","ocean_disable_title":"default","ocean_disable_heading":"default","ocean_post_title":"","ocean_post_subheading":"","ocean_post_title_style":"","ocean_post_title_background_color":"","ocean_post_title_background":0,"ocean_post_title_bg_image_position":"","ocean_post_title_bg_image_attachment":"","ocean_post_title_bg_image_repeat":"","ocean_post_title_bg_image_size":"","ocean_post_title_height":0,"ocean_post_title_bg_overlay":0.5,"ocean_post_title_bg_overlay_color":"","ocean_disable_breadcrumbs":"default","ocean_breadcrumbs_color":"","ocean_breadcrumbs_separator_color":"","ocean_breadcrumbs_links_color":"","ocean_breadcrumbs_links_hover_color":"","ocean_display_footer_widgets":"default","ocean_display_footer_bottom":"default","ocean_custom_footer_template":"","osh_disable_topbar_sticky":"default","osh_disable_header_sticky":"default","osh_sticky_header_style":"default","osh_sticky_header_effect":"","osh_custom_sticky_logo":0,"osh_custom_retina_sticky_logo":0,"osh_custom_sticky_logo_height":0,"osh_background_color":"","osh_links_color":"","osh_links_hover_color":"","osh_links_active_color":"","osh_links_bg_color":"","osh_links_hover_bg_color":"","osh_links_active_bg_color":"","osh_menu_social_links_color":"","osh_menu_social_hover_links_color":"","ocean_post_oembed":"","ocean_post_self_hosted_media":"","ocean_post_video_embed":"","ocean_link_format":"","ocean_link_format_target":"self","ocean_quote_format":"","ocean_quote_format_link":"post","ocean_gallery_link_images":"on","ocean_gallery_id":[],"footnotes":""},"categories":[16],"tags":[],"class_list":["post-2258","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","entry","has-media"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.3 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste - askem<\/title>\n<meta name=\"description\" content=\"ASKEM BUREAU D&#039;\u00c9TUDES ET DE FORMATION NUM\u00c9RIQUE. Nous vous assistons dans la transformation num\u00e9rique de vos outils, services et organisations tout en pla\u00e7ant l\u2019humain au c\u0153ur de notre d\u00e9marche d\u2019accompagnement.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/askem.eu\/en\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste - askem\" \/>\n<meta property=\"og:description\" content=\"ASKEM BUREAU D&#039;\u00c9TUDES ET DE FORMATION NUM\u00c9RIQUE. Nous vous assistons dans la transformation num\u00e9rique de vos outils, services et organisations tout en pla\u00e7ant l\u2019humain au c\u0153ur de notre d\u00e9marche d\u2019accompagnement.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/askem.eu\/en\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/\" \/>\n<meta property=\"og:site_name\" content=\"askem\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/fb.me\/askem.eu\" \/>\n<meta property=\"article:published_time\" content=\"2026-04-16T08:27:46+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-04-16T08:27:52+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:auto\/h:auto\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2026\/04\/sujet-askem-2026-04-16.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"800\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"askemadmin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"askemadmin\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/\"},\"author\":{\"name\":\"askemadmin\",\"@id\":\"https:\\\/\\\/askem.eu\\\/#\\\/schema\\\/person\\\/8bbee74ab9a977d56bf4826662e9d2e9\"},\"headline\":\"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste\",\"datePublished\":\"2026-04-16T08:27:46+00:00\",\"dateModified\":\"2026-04-16T08:27:52+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/\"},\"wordCount\":1207,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\/\\/askem.eu\\/wp-content\\/uploads\\/2026\\/04\\/sujet-askem-2026-04-16.png\",\"articleSection\":[\"AI\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/\",\"url\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/\",\"name\":\"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste - askem\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\/\\/askem.eu\\/wp-content\\/uploads\\/2026\\/04\\/sujet-askem-2026-04-16.png\",\"datePublished\":\"2026-04-16T08:27:46+00:00\",\"dateModified\":\"2026-04-16T08:27:52+00:00\",\"description\":\"ASKEM BUREAU D'\u00c9TUDES ET DE FORMATION NUM\u00c9RIQUE. Nous vous assistons dans la transformation num\u00e9rique de vos outils, services et organisations tout en pla\u00e7ant l\u2019humain au c\u0153ur de notre d\u00e9marche d\u2019accompagnement.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#primaryimage\",\"url\":\"https:\\/\\/askem.eu\\/wp-content\\/uploads\\/2026\\/04\\/sujet-askem-2026-04-16.png\",\"contentUrl\":\"https:\\/\\/askem.eu\\/wp-content\\/uploads\\/2026\\/04\\/sujet-askem-2026-04-16.png\",\"width\":1200,\"height\":800},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/askem.eu\\\/2026\\\/04\\\/16\\\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Accueil\",\"item\":\"https:\\\/\\\/askem.eu\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/askem.eu\\\/#website\",\"url\":\"https:\\\/\\\/askem.eu\\\/\",\"name\":\"askem\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/askem.eu\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/askem.eu\\\/#organization\",\"name\":\"Askem\",\"url\":\"https:\\\/\\\/askem.eu\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/askem.eu\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\/\\/mlpi0fxo3sth.i.optimole.com\\/cb:3obA.c61\\/w:760\\/h:480\\/q:mauto\\/f:best\\/https:\\/\\/askem.eu\\/wp-content\\/uploads\\/2020\\/10\\/logoGalaxieAskem3.png\",\"contentUrl\":\"https:\\/\\/mlpi0fxo3sth.i.optimole.com\\/cb:3obA.c61\\/w:760\\/h:480\\/q:mauto\\/f:best\\/https:\\/\\/askem.eu\\/wp-content\\/uploads\\/2020\\/10\\/logoGalaxieAskem3.png\",\"width\":760,\"height\":480,\"caption\":\"Askem\"},\"image\":{\"@id\":\"https:\\\/\\\/askem.eu\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/fb.me\\\/askem.eu\",\"https:\\\/\\\/linkedin.com\\\/company\\\/askem-eu\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/askem.eu\\\/#\\\/schema\\\/person\\\/8bbee74ab9a977d56bf4826662e9d2e9\",\"name\":\"askemadmin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a202f744ee3a4b6fdbe2ceb57fd84c72559337791a276662270d8d2fb7842e3f?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a202f744ee3a4b6fdbe2ceb57fd84c72559337791a276662270d8d2fb7842e3f?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a202f744ee3a4b6fdbe2ceb57fd84c72559337791a276662270d8d2fb7842e3f?s=96&d=mm&r=g\",\"caption\":\"askemadmin\"},\"sameAs\":[\"https:\\\/\\\/askem.eu\"]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste - askem","description":"ASKEM BUREAU D'\u00c9TUDES ET DE FORMATION NUM\u00c9RIQUE. Nous vous assistons dans la transformation num\u00e9rique de vos outils, services et organisations tout en pla\u00e7ant l\u2019humain au c\u0153ur de notre d\u00e9marche d\u2019accompagnement.","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:\/\/askem.eu\/en\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/","og_locale":"en_US","og_type":"article","og_title":"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste - askem","og_description":"ASKEM BUREAU D'\u00c9TUDES ET DE FORMATION NUM\u00c9RIQUE. Nous vous assistons dans la transformation num\u00e9rique de vos outils, services et organisations tout en pla\u00e7ant l\u2019humain au c\u0153ur de notre d\u00e9marche d\u2019accompagnement.","og_url":"https:\/\/askem.eu\/en\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/","og_site_name":"askem","article_publisher":"https:\/\/fb.me\/askem.eu","article_published_time":"2026-04-16T08:27:46+00:00","article_modified_time":"2026-04-16T08:27:52+00:00","og_image":[{"width":1200,"height":800,"url":"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:auto\/h:auto\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2026\/04\/sujet-askem-2026-04-16.png","type":"image\/png"}],"author":"askemadmin","twitter_card":"summary_large_image","twitter_misc":{"Written by":"askemadmin","Est. reading time":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#article","isPartOf":{"@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/"},"author":{"name":"askemadmin","@id":"https:\/\/askem.eu\/#\/schema\/person\/8bbee74ab9a977d56bf4826662e9d2e9"},"headline":"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste","datePublished":"2026-04-16T08:27:46+00:00","dateModified":"2026-04-16T08:27:52+00:00","mainEntityOfPage":{"@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/"},"wordCount":1207,"commentCount":0,"publisher":{"@id":"https:\/\/askem.eu\/#organization"},"image":{"@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#primaryimage"},"thumbnailUrl":"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:auto\/h:auto\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2026\/04\/sujet-askem-2026-04-16.png","articleSection":["AI"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/","url":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/","name":"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste - askem","isPartOf":{"@id":"https:\/\/askem.eu\/#website"},"primaryImageOfPage":{"@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#primaryimage"},"image":{"@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#primaryimage"},"thumbnailUrl":"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:auto\/h:auto\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2026\/04\/sujet-askem-2026-04-16.png","datePublished":"2026-04-16T08:27:46+00:00","dateModified":"2026-04-16T08:27:52+00:00","description":"ASKEM BUREAU D'\u00c9TUDES ET DE FORMATION NUM\u00c9RIQUE. Nous vous assistons dans la transformation num\u00e9rique de vos outils, services et organisations tout en pla\u00e7ant l\u2019humain au c\u0153ur de notre d\u00e9marche d\u2019accompagnement.","breadcrumb":{"@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#primaryimage","url":"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:auto\/h:auto\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2026\/04\/sujet-askem-2026-04-16.png","contentUrl":"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:auto\/h:auto\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2026\/04\/sujet-askem-2026-04-16.png","width":1200,"height":800},{"@type":"BreadcrumbList","@id":"https:\/\/askem.eu\/2026\/04\/16\/unsloth-fine-tuner-des-llm-open-source-rapidement-sur-du-materiel-modeste\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Accueil","item":"https:\/\/askem.eu\/"},{"@type":"ListItem","position":2,"name":"Unsloth : fine-tuner des LLM open source rapidement sur du mat\u00e9riel modeste"}]},{"@type":"WebSite","@id":"https:\/\/askem.eu\/#website","url":"https:\/\/askem.eu\/","name":"askem","description":"","publisher":{"@id":"https:\/\/askem.eu\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/askem.eu\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/askem.eu\/#organization","name":"Askem","url":"https:\/\/askem.eu\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/askem.eu\/#\/schema\/logo\/image\/","url":"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:760\/h:480\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2020\/10\/logoGalaxieAskem3.png","contentUrl":"https:\/\/mlpi0fxo3sth.i.optimole.com\/cb:3obA.c61\/w:760\/h:480\/q:mauto\/f:best\/https:\/\/askem.eu\/wp-content\/uploads\/2020\/10\/logoGalaxieAskem3.png","width":760,"height":480,"caption":"Askem"},"image":{"@id":"https:\/\/askem.eu\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/fb.me\/askem.eu","https:\/\/linkedin.com\/company\/askem-eu"]},{"@type":"Person","@id":"https:\/\/askem.eu\/#\/schema\/person\/8bbee74ab9a977d56bf4826662e9d2e9","name":"askemadmin","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/a202f744ee3a4b6fdbe2ceb57fd84c72559337791a276662270d8d2fb7842e3f?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/a202f744ee3a4b6fdbe2ceb57fd84c72559337791a276662270d8d2fb7842e3f?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/a202f744ee3a4b6fdbe2ceb57fd84c72559337791a276662270d8d2fb7842e3f?s=96&d=mm&r=g","caption":"askemadmin"},"sameAs":["https:\/\/askem.eu"]}]}},"_links":{"self":[{"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/posts\/2258","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/comments?post=2258"}],"version-history":[{"count":1,"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/posts\/2258\/revisions"}],"predecessor-version":[{"id":2260,"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/posts\/2258\/revisions\/2260"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/media\/2259"}],"wp:attachment":[{"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/media?parent=2258"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/categories?post=2258"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/askem.eu\/en\/wp-json\/wp\/v2\/tags?post=2258"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}