{"id":5661,"date":"2024-07-23T13:30:44","date_gmt":"2024-07-23T04:30:44","guid":{"rendered":"https:\/\/highreso.jp\/edgehub\/?p=5661"},"modified":"2024-07-23T13:56:31","modified_gmt":"2024-07-23T04:56:31","slug":"llama3torchtune","status":"publish","type":"post","link":"https:\/\/highreso.jp\/edgehub\/machinelearning\/llama3torchtune.html","title":{"rendered":"\u3010Llama3\u3011Torchtune\u3067\u7c21\u5358\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0(QLoRA)"},"content":{"rendered":"\n<p><span class=\"swl-format-2\">\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001PyTorch\u306eTorchtune\u3092\u4f7f\u3063\u305fLlama3\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0(QLoRA)\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<p>Torchtune\u3092\u4f7f\u3063\u3066\u7c21\u5358\u306b\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u5b9f\u88c5\u304c\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Torchtune\u3067Llama3\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\uff08QLoRA\uff09<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"600\" height=\"300\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/atom_midashi.jpg\" alt=\"\u898b\u51fa\u3057\u753b\u50cf\" class=\"wp-image-5357\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/atom_midashi.jpg 600w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/atom_midashi-300x150.jpg 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">Torchtune\u3068\u306f<\/h3>\n\n\n\n<p>\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001PyTorch\u306eTorchtune\u3092\u4f7f\u3063\u3066\u3001Llama3\u306e\u30e2\u30c7\u30eb\u306bQLoRA\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">Torchtune\u306f\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u7c21\u5358\u306b\u5b9f\u884c\u3059\u308b\u305f\u3081\u306ePyTorch\u30cd\u30a4\u30c6\u30a3\u30d6\u306e\u30e9\u30a4\u30d6\u30e9\u30ea<\/span>\u3067\u3001\u4ee5\u4e0b\u306e\u3088\u3046\u306a\u7279\u5fb4\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<div class=\"wp-block-group is-style-big_icon_point\"><div class=\"wp-block-group__inner-container\">\n<ul class=\"wp-block-list\">\n<li>LoRA\u3001QLoRA\u3001\u30d5\u30eb\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306a\u3069\u8907\u6570\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u65b9\u6cd5\u3092\u63d0\u4f9b<\/li>\n\n\n\n<li>\u30b7\u30f3\u30b0\u30ebGPU\u306b\u3088\u308b\u5b66\u7fd2\u304b\u3089\u3001\u30de\u30eb\u30c1GPU\u306b\u3088\u308b\u5206\u6563\u5b66\u7fd2\u307e\u3067\u306e\u65b9\u6cd5\u3092\u63d0\u4f9b<\/li>\n\n\n\n<li>\u3042\u3089\u3086\u308b\u7a2e\u985e\u306e\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u304c\u7528\u610f\u3055\u308c\u3066\u304a\u308a\u3001\u7c21\u5358\u306b\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u8a66\u305b\u308b<\/li>\n\n\n\n<li>YAML\u5f62\u5f0f\u3067\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3001\u63a8\u8ad6\u306b\u4f7f\u7528\u3059\u308b\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u6307\u5b9a\u3057\u3066\u5b9f\u884c<\/li>\n\n\n\n<li>HuggingFace\u3001bitsandbytes\u3001WandB\u306a\u3069\u306e\u5916\u90e8\u306e\u30e9\u30a4\u30d6\u30e9\u30ea\u3084\u30c4\u30fc\u30eb\u304c\u5229\u7528\u3067\u304d\u308b<\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<div class=\"swell-block-capbox cap_box is-style-onborder_ttl2\" data-colset=\"col3\"><div class=\"cap_box_ttl\"><span>\u7528\u8a9e\u306e\u89e3\u8aac<\/span><\/div><div class=\"cap_box_content\">\n<p><span class=\"swl-format-2\">QLoRA<\/span><\/p>\n\n\n\n<p>QLoRA\u306f\u3001LoRA\u3068\u91cf\u5b50\u5316\uff08Quantization\uff09\u306e\uff12\u3064\u306e\u8981\u7d20\u3092\u3082\u3064\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u624b\u6cd5\u3067\u3059\u3002<\/p>\n\n\n\n<p>LoRA\u306f\u3001\u30e2\u30c7\u30eb\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u4f4e\u30e9\u30f3\u30af\u884c\u5217\u3067\u8fd1\u4f3c\u3059\u308b\u3053\u3068\u3067\u3001\u66f4\u65b0\u3059\u308b\u30d1\u30e9\u30e1\u30fc\u30bf\u6570\u3092\u5927\u5e45\u306b\u6e1b\u3089\u3057\u8a08\u7b97\u91cf\u3092\u524a\u6e1b\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u91cf\u5b50\u5316\u3068\u306f\u3001\u30e2\u30c7\u30eb\u306e\u7cbe\u5ea6\u3092\u4e0b\u3052\u308b\u4ee3\u308f\u308a\u306b\u3001GPU\u30e1\u30e2\u30ea\u3092\u5927\u5e45\u306b\u7bc0\u7d04\u3059\u308b\u6280\u8853\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n<\/div><\/div>\n\n\n\n<p class=\"is-style-balloon_box\">Llama3\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u77e5\u308a\u305f\u3044\u65b9\u306f\u3001\u4ee5\u4e0b\u306e\u8a18\u4e8b\u3092\u3054\u89a7\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n<div class=\"swell-block-postLink\">\t\t\t<div class=\"p-blogCard -internal\" data-type=\"type1\" data-onclick=\"clickLink\">\n\t\t\t\t<div class=\"p-blogCard__inner\">\n\t\t\t\t\t<span class=\"p-blogCard__caption\">\u3042\u308f\u305b\u3066\u8aad\u307f\u305f\u3044<\/span>\n\t\t\t\t\t<div class=\"p-blogCard__thumb c-postThumb\"><figure class=\"c-postThumb__figure\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3toha-eyecatch-300x158.jpg\" alt=\"\" class=\"c-postThumb__img u-obf-cover\" width=\"320\" height=\"180\"><\/figure><\/div>\t\t\t\t\t<div class=\"p-blogCard__body\">\n\t\t\t\t\t\t<a class=\"p-blogCard__title\" href=\"https:\/\/highreso.jp\/edgehub\/machinelearning\/llama3-howto.html\">Llama3\u3068\u306f\uff1f\u4f7f\u3044\u65b9\u304b\u3089\u6027\u80fd\u3001\u5546\u7528\u5229\u7528\u307e\u3067\u5206\u304b\u308a\u3084\u3059\u304f\u89e3\u8aac\uff01<\/a>\n\t\t\t\t\t\t<span class=\"p-blogCard__excerpt\">Llama3\u306f\u3001Meta\u304c\u958b\u767a\u3057\u305f\u7121\u6599\u3067\u4f7f\u3048\u308b\u9ad8\u6027\u80fdLLM\u3067\u3059\u3002 Llama3 70B\u306e\u6027\u80fd\u306f\u300cGemini1.5Pro\u300d\u3084\u300cClaude3 Sonnet\u300d\u3092\u4e0a\u56de\u308b\u3068\u8a00\u308f\u308c\u3066\u3044\u307e\u3059\u3002 \u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001Llama3\u306e\u6027&#8230;<\/span>\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\n\n<p class=\"is-style-balloon_box\">HuggingFace\u306eSFTTrainer\u3067\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u65b9\u6cd5\u306f\u5225\u8a18\u4e8b\u3067\u89e3\u8aac\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n<div class=\"swell-block-postLink\">\t\t\t<div class=\"p-blogCard -internal\" data-type=\"type1\" data-onclick=\"clickLink\">\n\t\t\t\t<div class=\"p-blogCard__inner\">\n\t\t\t\t\t<span class=\"p-blogCard__caption\">\u3042\u308f\u305b\u3066\u8aad\u307f\u305f\u3044<\/span>\n\t\t\t\t\t<div class=\"p-blogCard__thumb c-postThumb\"><figure class=\"c-postThumb__figure\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3sftqlora-eyecatch-300x158.jpg\" alt=\"\" class=\"c-postThumb__img u-obf-cover\" width=\"320\" height=\"180\"><\/figure><\/div>\t\t\t\t\t<div class=\"p-blogCard__body\">\n\t\t\t\t\t\t<a class=\"p-blogCard__title\" href=\"https:\/\/highreso.jp\/edgehub\/machinelearning\/llama3sftqlora.html\">\u3010Llama3\u3011SFTTrainer\u3067\u7c21\u5358\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0(QLoRA)<\/a>\n\t\t\t\t\t\t<span class=\"p-blogCard__excerpt\">\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001Hugging Face\u306eSFTTrainer\u3092\u4f7f\u3063\u305fLlama3\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0(QLoRA)\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002 SFTTrainer\u3092\u4f7f\u3046\u3053\u3068\u3067\u3001\u5c11\u306a\u3044\u30b3\u30fc\u30c9\u3067\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e&#8230;<\/span>\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\n\n<p class=\"is-style-balloon_box\">Unsloth\u3092\u4f7f\u3063\u305f\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306b\u3064\u3044\u3066\u306f\u5225\u8a18\u4e8b\u3067\u89e3\u8aac\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n<div class=\"swell-block-postLink\">\t\t\t<div class=\"p-blogCard -internal\" data-type=\"type1\" data-onclick=\"clickLink\">\n\t\t\t\t<div class=\"p-blogCard__inner\">\n\t\t\t\t\t<span class=\"p-blogCard__caption\">\u3042\u308f\u305b\u3066\u8aad\u307f\u305f\u3044<\/span>\n\t\t\t\t\t<div class=\"p-blogCard__thumb c-postThumb\"><figure class=\"c-postThumb__figure\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3unsloth-eyecatch-300x158.jpg\" alt=\"\" class=\"c-postThumb__img u-obf-cover\" width=\"320\" height=\"180\"><\/figure><\/div>\t\t\t\t\t<div class=\"p-blogCard__body\">\n\t\t\t\t\t\t<a class=\"p-blogCard__title\" href=\"https:\/\/highreso.jp\/edgehub\/machinelearning\/llama3unsloth.html\">\u3010Llama3\u3011Unsloth\u3067\u7206\u901f\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0(QLoRA)<\/a>\n\t\t\t\t\t\t<span class=\"p-blogCard__excerpt\">\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001Unsloth\u3092\u4f7f\u3063\u305fLlama3\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0(QLoRA)\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002 Unsloth\u306b\u3088\u308aGPU\u30e1\u30e2\u30ea\u306e\u4f7f\u7528\u91cf\u3092\u5927\u5e45\u306b\u524a\u6e1b\u3057\u3066\u3001\u9ad8\u901f\u306b\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u304c\u3067&#8230;<\/span>\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\n\n<h3 class=\"wp-block-heading\">\u4f7f\u7528\u3059\u308b\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8<\/h3>\n\n\n\n<p>\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u306b\u306f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u300cbbz662bbz\/databricks-dolly-15k-ja-gozaru\u300d\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>15,000\u4ee5\u4e0a\u306e\u6307\u793a\u3068\u5fdc\u7b54\u3067\u69cb\u6210\u3055\u308c\u305f\u65e5\u672c\u8a9e\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3067\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">\u5fdc\u7b54\u306e\u8a9e\u5c3e\u304c\u300c\u3054\u3056\u308b\u300d\u306e\u53e3\u8abf\u306b\u306a\u3063\u3066\u3044\u308b\u3053\u3068\u304c\u7279\u5fb4\u3067\u3059\u3002<\/span><\/p>\n\n\n\n<figure class=\"wp-block-embed\"><div class=\"wp-block-embed__wrapper\">\nhttps:\/\/huggingface.co\/datasets\/bbz662bbz\/databricks-dolly-15k-ja-gozaru\n<\/div><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">\u4e8b\u524d\u6e96\u5099<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"600\" height=\"300\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/woodtexture_midashi.jpg\" alt=\"\u898b\u51fa\u3057\u753b\u50cf\" class=\"wp-image-5514\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/woodtexture_midashi.jpg 600w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/woodtexture_midashi-300x150.jpg 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">\u5fc5\u8981\u306a\u30b9\u30da\u30c3\u30af\u30fb\u5b9f\u884c\u74b0\u5883<\/h3>\n\n\n\n<p>Llama3\u306eQLoRA\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3067\u306f\u3001\u5927\u5bb9\u91cf\u306eGPU\u30e1\u30e2\u30ea\u3092\u5fc5\u8981\u3068\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001<span class=\"swl-format-2\">GPU\u30e1\u30e2\u30ea80GB\u3092\u642d\u8f09\u3057\u305fNVIDIA A100 80GB\u306e\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3092\u4f7f\u7528\u3057\u3066\u3044\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<p>\u5b9f\u884c\u74b0\u5883\u306e\u8a73\u7d30\u306f\u4ee5\u4e0b\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n\n\n\n<div class=\"wp-block-group is-style-big_icon_good\"><div class=\"wp-block-group__inner-container\">\n<ul class=\"wp-block-list\">\n<li>GPU\uff1aNVIDIA A100 80GB<\/li>\n\n\n\n<li>GPU\u30e1\u30e2\u30ea\uff08VRAM\uff09\uff1a80GB<\/li>\n\n\n\n<li>OS\uff1aUbuntu22.04<\/li>\n\n\n\n<li>Docker<\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<p>GPU\u30e1\u30e2\u30ea\u306e\u4f7f\u7528\u91cf\u306f\u3001Github\u304c\u53c2\u8003\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<figure class=\"wp-block-embed\"><div class=\"wp-block-embed__wrapper\">\nhttps:\/\/github.com\/pytorch\/torchtune?tab=readme-ov-file#fine-tuning-recipes\n<\/div><\/figure>\n\n\n\n<div class=\"p-blogParts post_content\" data-partsID=\"4485\"><p><div class=\"p-adBox -ranking -border-on\" data-id=\"4540\" data-ad=\"ranking\"><div class=\"p-adBox__title -rank1\">50\uff05\u5b89\u3044\u696d\u754c\u6700\u5b89\u7d1aGPU\u30af\u30e9\u30a6\u30c9\u300cGPUSOROBAN\u300d<\/div><div class=\"p-adBox__body\"><div class=\"p-adBox__img\"><div class=\"centerw250\"><a href=\"https:\/\/soroban.highreso.jp\/?utm_medium=edgehub&utm_source=edgehub&utm_campaign=highend-boxtop&adid=edgehub\" rel=\"nofollow\" referrerpolicy=\"no-referrer-when-downgrade\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/gpusorobanlllm_banner.jpg\" alt=\"GPUSOROBAN\"><\/a><\/div><\/div><div class=\"p-adBox__details\"><div class=\"p-adBox__star c-reviewStars\"><i class=\"icon-star-full\"><\/i><i class=\"icon-star-full\"><\/i><i class=\"icon-star-full\"><\/i><i class=\"icon-star-full\"><\/i><i class=\"icon-star-full\"><\/i><\/div><div class=\"p-adBox__price u-thin u-fz-s\">GPU\u306e\u30b3\u30b9\u30c8\u3067\u304a\u60a9\u307f\u306e\u65b9\u306b\uff01<\/div><div class=\"p-adBox__desc\">AWS\u3068\u6bd4\u3079\u306650%\u4ee5\u4e0a\u5b89\u3044\u6599\u91d1\u3067\u3001\u9ad8\u901fGPU\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3092\u3054\u5229\u7528\u3044\u305f\u3060\u3051\u307e\u3059\u3002\r\n\u9762\u5012\u306a\u8a2d\u5b9a\u306f\u4e0d\u8981\u3001\u308f\u305a\u304b3\u5206\u3067\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u306e\u7acb\u3061\u4e0a\u3052\u304c\u53ef\u80fd\u3067\u3059\u3002\r\n\u30bf\u30b9\u30af\u306b\u5fdc\u3058\u3066\u6700\u9069\u306aGPU\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3092\u9078\u629e\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/div><\/div><\/div><div class=\"p-adBox__btns\"><a href=\"https:\/\/soroban.highreso.jp\/compute\/?utm_medium=edgehub&utm_source=edgehub&utm_campaign=boxtop&adid=edgehub\" class=\"p-adBox__btn -btn1\" target=\"_blank\" rel=\"noopener nofollow\">\u516c\u5f0f\u30b5\u30a4\u30c8\u3092\u898b\u308b<\/a><a href=\"https:\/\/soroban.highreso.jp\/download\/compute-catalog\/?utm_medium=edgehub&utm_source=edgehub&utm_campaign=boxtop&adid=edgehub\" class=\"p-adBox__btn -btn2\" target=\"_blank\" rel=\"noopener nofollow\">\u8cc7\u6599\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9<\/a><\/div><\/div><\/p>\n<\/div>\n\n\n\n<h3 class=\"wp-block-heading\">Llama3\u306e\u30e2\u30c7\u30eb\u5229\u7528\u7533\u8acb<\/h3>\n\n\n\n<p><span class=\"swl-format-2\">Llama3\u306e\u30e2\u30c7\u30eb\u3092\u4f7f\u3046\u306b\u3042\u305f\u3063\u3066\u3001\u5229\u7528\u7533\u8acb\u304c\u5fc5\u8981\u306b\u306a\u308a\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<p>\u4ee5\u4e0b\u306e\u8a18\u4e8b\u3067\u5229\u7528\u7533\u8acb\u306e\u65b9\u6cd5\u3092\u7d39\u4ecb\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n<div class=\"swell-block-postLink\">\t\t\t<div class=\"p-blogCard -internal\" data-type=\"type1\" data-onclick=\"clickLink\">\n\t\t\t\t<div class=\"p-blogCard__inner\">\n\t\t\t\t\t<span class=\"p-blogCard__caption\">\u3042\u308f\u305b\u3066\u8aad\u307f\u305f\u3044<\/span>\n\t\t\t\t\t<div class=\"p-blogCard__thumb c-postThumb\"><figure class=\"c-postThumb__figure\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3toha-eyecatch-300x158.jpg\" alt=\"\" class=\"c-postThumb__img u-obf-cover\" width=\"320\" height=\"180\"><\/figure><\/div>\t\t\t\t\t<div class=\"p-blogCard__body\">\n\t\t\t\t\t\t<a class=\"p-blogCard__title\" href=\"https:\/\/highreso.jp\/edgehub\/machinelearning\/llama3-howto.html\">Llama3\u3068\u306f\uff1f\u4f7f\u3044\u65b9\u304b\u3089\u6027\u80fd\u3001\u5546\u7528\u5229\u7528\u307e\u3067\u5206\u304b\u308a\u3084\u3059\u304f\u89e3\u8aac\uff01<\/a>\n\t\t\t\t\t\t<span class=\"p-blogCard__excerpt\">Llama3\u306f\u3001Meta\u304c\u958b\u767a\u3057\u305f\u7121\u6599\u3067\u4f7f\u3048\u308b\u9ad8\u6027\u80fdLLM\u3067\u3059\u3002 Llama3 70B\u306e\u6027\u80fd\u306f\u300cGemini1.5Pro\u300d\u3084\u300cClaude3 Sonnet\u300d\u3092\u4e0a\u56de\u308b\u3068\u8a00\u308f\u308c\u3066\u3044\u307e\u3059\u3002 \u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001Llama3\u306e\u6027&#8230;<\/span>\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\n\n<h2 class=\"wp-block-heading\">Docker\u3067\u74b0\u5883\u69cb\u7bc9<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"600\" height=\"300\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/moutainsunset_midashi.jpg\" alt=\"\u898b\u51fa\u3057\u753b\u50cf\" class=\"wp-image-5173\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/moutainsunset_midashi.jpg 600w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/moutainsunset_midashi-300x150.jpg 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n\n\n\n<p><span class=\"swl-format-2\">Docker\u3092\u4f7f\u7528\u3057\u3066Llama3\u306e\u74b0\u5883\u69cb\u7bc9\u3092\u3057\u3066\u3044\u304d\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<p>Docker\u306e\u4f7f\u3044\u65b9\u306f\u4ee5\u4e0b\u306e\u8a18\u4e8b\u3092\u3054\u89a7\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n<div class=\"swell-block-postLink\">\t\t\t<div class=\"p-blogCard -internal\" data-type=\"type1\" data-onclick=\"clickLink\">\n\t\t\t\t<div class=\"p-blogCard__inner\">\n\t\t\t\t\t<span class=\"p-blogCard__caption\">\u3042\u308f\u305b\u3066\u8aad\u307f\u305f\u3044<\/span>\n\t\t\t\t\t<div class=\"p-blogCard__thumb c-postThumb\"><figure class=\"c-postThumb__figure\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/ubuntudocker-eyecatch-300x158.jpg\" alt=\"\" class=\"c-postThumb__img u-obf-cover\" width=\"320\" height=\"180\"><\/figure><\/div>\t\t\t\t\t<div class=\"p-blogCard__body\">\n\t\t\t\t\t\t<a class=\"p-blogCard__title\" href=\"https:\/\/highreso.jp\/edgehub\/machinelearning\/ubuntudocker.html\">Ubuntu\u306bDocker\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\uff01GPU\u3092\u30b3\u30f3\u30c6\u30ca\u304b\u3089\u4f7f\u3046\u65b9\u6cd5<\/a>\n\t\t\t\t\t\t<span class=\"p-blogCard__excerpt\">GPU\u306e\u74b0\u5883\u69cb\u7bc9\u3067\u306f\u3001\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u4e0d\u4e00\u81f4\u306b\u3088\u308b\u30a8\u30e9\u30fc\u304c\u983b\u767a\u3057\u307e\u3059\u3002 Docker\u3092\u4f7f\u3046\u3053\u3068\u3067\u3001\u3053\u306e\u3088\u3046\u306a\u4f9d\u5b58\u95a2\u4fc2\u306e\u30a8\u30e9\u30fc\u3092\u56de\u907f\u3057\u3084\u3059\u304f\u306a\u308a\u307e\u3059\u3002 \u3053\u306e\u8a18\u4e8b\u3067&#8230;<\/span>\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\n\n<div class=\"swell-block-step\" data-num-style=\"circle\">\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">Dockerfile\u306e\u4f5c\u6210<\/div><div class=\"swell-block-step__body\">\n<p><span class=\"swl-format-2\">Dokcerfile\u306b\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u30d1\u30c3\u30b1\u30fc\u30b8\u3092\u8a18\u8ff0\u3057\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<div class=\"swell-block-capbox cap_box is-style-onborder_ttl2\" data-colset=\"col3\"><div class=\"cap_box_ttl\"><span>\u30d1\u30c3\u30b1\u30fc\u30b8\u4e00\u89a7<\/span><\/div><div class=\"cap_box_content\">\n<ul class=\"wp-block-list\">\n<li>CUDA\uff1a12.1<\/li>\n\n\n\n<li>Python\uff1a3.10<\/li>\n\n\n\n<li>PyTorch\uff1a2.2.2<\/li>\n\n\n\n<li>Torchtune\uff1a0.2.0.dev20240607+cu121<\/li>\n\n\n\n<li>WandB<\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<p class=\"is-style-icon_info\">Torchtune\u3001CUDA\u3001PyTorch\u7b49\u306f\u3001\u30d0\u30fc\u30b8\u30e7\u30f3\u4f9d\u5b58\u95a2\u4fc2\u306b\u3088\u308b\u30a8\u30e9\u30fc\u304c\u8d77\u304d\u3084\u3059\u3044\u306e\u3067\u3001\u52d5\u4f5c\u691c\u8a3c\u6e08\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u6307\u5b9a\u3057\u3066\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>Ubuntu\u306e\u30b3\u30de\u30f3\u30c9\u30e9\u30a4\u30f3\u304b\u3089\u3001Dockerfile\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>mkdir llama3_torchtune\ncd llama3_torchtune\nnano Dockerfile<\/code><\/pre>\n\n\n\n<p>\u6b21\u306e\u8a18\u8ff0\u3092\u30b3\u30d4\u30fc\u3057\u3066Dokcerfile\u306b\u8cbc\u308a\u4ed8\u3051\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code># \u30d9\u30fc\u30b9\u30a4\u30e1\u30fc\u30b8(CUDA)\u306e\u6307\u5b9a\nFROM nvidia\/cuda:12.1.0-cudnn8-devel-ubuntu22.04\n\n# \u5fc5\u8981\u306a\u30d1\u30c3\u30b1\u30fc\u30b8\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\nRUN apt-get update &amp;&amp; apt-get install -y python3-pip python3-venv git nano\n\n# \u4f5c\u696d\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3092\u8a2d\u5b9a\nWORKDIR \/app\n\n# \u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\nCOPY . \/app\n\n# Python\u4eee\u60f3\u74b0\u5883\u306e\u4f5c\u6210\nRUN python3 -m venv \/app\/.venv\n\n# \u4eee\u60f3\u74b0\u5883\u3092\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30c8\u3059\u308b\u30b3\u30de\u30f3\u30c9\u3092.bashrc\u306b\u8ffd\u52a0\nRUN echo \"source \/app\/.venv\/bin\/activate\" &gt;&gt; \/root\/.bashrc\n\n# PyTorch\u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\nRUN \/app\/.venv\/bin\/pip install torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 --index-url https:\/\/download.pytorch.org\/whl\/cu121\n\n# Torchtune\u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\nRUN \/app\/.venv\/bin\/pip install --pre torchtune --extra-index-url https:\/\/download.pytorch.org\/whl\/nightly\/cu121\n\n# Wandb\u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\nRUN \/app\/.venv\/bin\/pip install wandb\n\n# \u30b3\u30f3\u30c6\u30ca\u306e\u8d77\u52d5\u6642\u306bbash\u3092\u5b9f\u884c\nCMD &#091;\"\/bin\/bash\"]<\/code><\/pre>\n\n\n\n<p class=\"is-style-icon_announce\">[Ctrl + S]\u30ad\u30fc\u3067\u5909\u66f4\u5185\u5bb9\u3092\u4fdd\u5b58\u3057\u3001[Ctrl + X]\u30ad\u30fc\u3067\u7de8\u96c6\u30e2\u30fc\u30c9\u304b\u3089\u629c\u3051\u307e\u3059\u3002<\/p>\n<\/div><\/div>\n\n\n\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">docker-compose.yml\u30d5\u30a1\u30a4\u30eb\u306e\u4f5c\u6210<\/div><div class=\"swell-block-step__body\">\n<p><span class=\"swl-format-2\">docker-compose.yml\u30d5\u30a1\u30a4\u30eb\u3092\u4f7f\u3063\u3066Docker\u30b3\u30f3\u30c6\u30ca\u306e\u8a2d\u5b9a\u3092\u3057\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<p>docker-compose.yml\u30d5\u30a1\u30a4\u30eb\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>nano docker-compose.yml<\/code><\/pre>\n\n\n\n<p>\u6b21\u306e\u8a18\u8ff0\u3092\u30b3\u30d4\u30fc\u3057\u3066docker-compose.yml\u306b\u8cbc\u308a\u4ed8\u3051\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>services:\n  llama3_torchtune:\n    build:\n      context: .\n      dockerfile: Dockerfile\n    image: llama3_torchtune\n    runtime: nvidia\n    container_name: llama3_torchtune\n    volumes:\n      - .:\/app\/torchtune\n    deploy:\n      resources:\n        reservations:\n          devices:\n            - driver: nvidia\n              count: 1\n              capabilities: &#091;gpu]<\/code><\/pre>\n\n\n\n<p class=\"is-style-icon_announce\">[Ctrl + S]\u30ad\u30fc\u3067\u5909\u66f4\u5185\u5bb9\u3092\u4fdd\u5b58\u3057\u3001[Ctrl + X]\u30ad\u30fc\u3067\u7de8\u96c6\u30e2\u30fc\u30c9\u304b\u3089\u629c\u3051\u307e\u3059\u3002<\/p>\n<\/div><\/div>\n\n\n\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">Docker\u30b3\u30f3\u30c6\u30ca\u3092\u8d77\u52d5<\/div><div class=\"swell-block-step__body\">\n<p><span class=\"swl-format-2\">Dockerfile\u304b\u3089\u30d3\u30eb\u30c9\u3057\u3066\u30b3\u30f3\u30c6\u30ca\u3092\u8d77\u52d5\u3057\u307e\u3059\u3002&nbsp;&nbsp;&nbsp;<\/span><\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>docker compose run llama3_torchtune<\/code><\/pre>\n<\/div><\/div>\n<\/div>\n\n\n\n<h2 class=\"wp-block-heading\">Llama3\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u5b9f\u88c5<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"600\" height=\"300\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/goldpaint_midashi.jpg\" alt=\"\u898b\u51fa\u3057\u753b\u50cf\" class=\"wp-image-5504\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/goldpaint_midashi.jpg 600w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/goldpaint_midashi-300x150.jpg 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n\n\n\n<p><span class=\"swl-format-2\">\u8d77\u52d5\u3057\u305fDokcer\u30b3\u30f3\u30c6\u30ca\u4e0a\u3067\u3001Llama3\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u5b9f\u88c5\u3057\u3066\u3044\u304d\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<div class=\"swell-block-step\" data-num-style=\"circle\">\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">\u30e2\u30c7\u30eb\u3068\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u306e\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9<\/div><div class=\"swell-block-step__body\">\n<p><span class=\"swl-format-2\">\u30e2\u30c7\u30eb\u3068\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3057\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>tune download meta-llama\/Meta-Llama-3-8B-Instruct \\\n--output-dir \/app\/torchtune\/models \\\n--hf-token '************'<\/code><\/pre>\n\n\n\n<div class=\"swell-block-capbox cap_box is-style-onborder_ttl2\" data-colset=\"col3\"><div class=\"cap_box_ttl\"><span>\u30b3\u30fc\u30c9\u306e\u89e3\u8aac<\/span><\/div><div class=\"cap_box_content\">\n<p><span class=\"swl-format-2\">meta-llama\/Meta-Llama-3-8B-Instruct<\/span><\/p>\n\n\n\n<p>\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3059\u308b\u30e2\u30c7\u30ebID\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">&#8211;output-dir<\/span><\/p>\n\n\n\n<p>\u30e2\u30c7\u30eb\u306e\u4fdd\u5b58\u5148\u306e\u30d1\u30b9\u3092\u6307\u5b9a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">&#8211;hf-token<\/span><\/p>\n\n\n\n<p>Hugging Face\u306e\u30a2\u30af\u30bb\u30b9\u30c8\u30fc\u30af\u30f3\u304c\u5165\u308a\u307e\u3059\u3002<\/p>\n<\/div><\/div>\n\n\n\n<p>Hugging Face\u306e\u30a2\u30af\u30bb\u30b9\u30c8\u30fc\u30af\u30f3\u4f5c\u6210\u65b9\u6cd5\u306f\u4ee5\u4e0b\u306e\u8a18\u4e8b\u3067\u89e3\u8aac\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n<div class=\"swell-block-postLink\">\t\t\t<div class=\"p-blogCard -internal\" data-type=\"type1\" data-onclick=\"clickLink\">\n\t\t\t\t<div class=\"p-blogCard__inner\">\n\t\t\t\t\t<span class=\"p-blogCard__caption\">\u3042\u308f\u305b\u3066\u8aad\u307f\u305f\u3044<\/span>\n\t\t\t\t\t<div class=\"p-blogCard__thumb c-postThumb\"><figure class=\"c-postThumb__figure\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/huggingface-eyecatch-300x158.jpg\" alt=\"\" class=\"c-postThumb__img u-obf-cover\" width=\"320\" height=\"180\"><\/figure><\/div>\t\t\t\t\t<div class=\"p-blogCard__body\">\n\t\t\t\t\t\t<a class=\"p-blogCard__title\" href=\"https:\/\/highreso.jp\/edgehub\/machinelearning\/huggingfacetoken.html\">Hugging Face\u306e\u4f7f\u3044\u65b9\uff01\u30a2\u30af\u30bb\u30b9\u30c8\u30fc\u30af\u30f3\u4f5c\u6210\u304b\u3089\u30ed\u30b0\u30a4\u30f3\u307e\u3067<\/a>\n\t\t\t\t\t\t<span class=\"p-blogCard__excerpt\">\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001Hugging Face\u30a2\u30ab\u30a6\u30f3\u30c8\u306e\u4f5c\u308a\u65b9\u3084\u30b3\u30de\u30f3\u30c9\u30e9\u30a4\u30f3\u304b\u3089\u306e\u30ed\u30b0\u30a4\u30f3\u65b9\u6cd5\u3092\u7d39\u4ecb\u3057\u3066\u3044\u307e\u3059\u3002 Hugging Face\u3092\u4f7f\u3046\u3053\u3068\u3067\u3001\u751f\u6210AI\u306b\u5fc5\u8981\u306a\u30e2\u30c7\u30eb\u3084\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8&#8230;<\/span>\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div><\/div><\/div>\n\n\n\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u524d\u306e\u30e2\u30c7\u30eb\u3067\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u30c6\u30b9\u30c8<\/div><div class=\"swell-block-step__body\">\n<p><span class=\"swl-format-2\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u3059\u308b\u524d\u306e\u30e2\u30c7\u30eb\u3067\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u306e\u30c6\u30b9\u30c8\u3092\u3057\u3066\u307f\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<p>\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u306e\u63a8\u8ad6\u306b\u4f7f\u7528\u3059\u308byaml\u30d5\u30a1\u30a4\u30eb\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>nano torchtune\/inference_confg.yaml<\/code><\/pre>\n\n\n\n\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code># Model Arguments\nmodel:\n  _component_: torchtune.models.llama3.llama3_8b\n\ncheckpointer:\n  _component_: torchtune.utils.FullModelMetaCheckpointer\n  checkpoint_dir: \/app\/torchtune\/models\n  checkpoint_files: &#091;\/app\/torchtune\/models\/original\/consolidated.00.pth]\n  recipe_checkpoint: null\n  output_dir: \/app\/torchtune\/output\n  model_type: LLAMA3\n\ndevice: cuda\ndtype: bf16\nseed: null\n\n# Tokenizer arguments\ntokenizer:\n  _component_: torchtune.models.llama3.llama3_tokenizer\n  path: \/app\/torchtune\/models\/original\/tokenizer.model\n\n# Generation arguments\nprompt: \"AirPods\u3068\u306f\u4f55\u3067\u3059\u304b\"\nmax_new_tokens: 128\ntemperature: 0.8\ntop_k: 40\nstop_tokens: \"&lt;|eot_id|&gt;\"\n\nquantizer: null<\/code><\/pre>\n\n\n\n<p class=\"is-style-icon_announce\">[Ctrl + S]\u30ad\u30fc\u3067\u5909\u66f4\u5185\u5bb9\u3092\u4fdd\u5b58\u3057\u3001[Ctrl + X]\u30ad\u30fc\u3067\u7de8\u96c6\u30e2\u30fc\u30c9\u304b\u3089\u629c\u3051\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u306e\u30c6\u30b9\u30c8\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>tune run generate --config torchtune\/inference_confg.yaml<\/code><\/pre>\n\n\n<div class=\"swell-block-balloon\"><div class=\"c-balloon -bln-right\" data-col=\"green\"><div class=\"c-balloon__body -speaking -border-none\"><div class=\"c-balloon__text\">\n<p>AirPods\u3068\u306f\u4f55\u3067\u3059\u304b\uff1f<\/p>\n<span class=\"c-balloon__shapes\"><span class=\"c-balloon__before\"><\/span><span class=\"c-balloon__after\"><\/span><\/span><\/div><\/div><\/div><\/div>\n\n<div class=\"swell-block-balloon\"><div class=\"c-balloon -bln-left\" data-col=\"gray\"><div class=\"c-balloon__body -speaking -border-none\"><div class=\"c-balloon__text\">\n<p>AirPods is a brand of wireless earbuds developed by Apple Inc. They were first released in 2016 and have since become one of the most popular and widely used earbuds on the market.<br>AirPods are designed to be a convenient and easy-to-use alternative to traditional earbuds. They are powered by a rechargeable lithium-ion battery and have a wireless range of up to 30 feet (9 meters). They are also sweat and water resistant, making them a great option for athletes and fitness enthusiasts.<br>One of the key features of AirPods is their ability to seamlessly connect to Apple devices such as<\/p>\n<span class=\"c-balloon__shapes\"><span class=\"c-balloon__before\"><\/span><span class=\"c-balloon__after\"><\/span><\/span><\/div><\/div><\/div><\/div>\n\n\n<p class=\"is-style-icon_announce\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u524d\u306e\u30e2\u30c7\u30eb\u3067\u306f\u3001\u82f1\u8a9e\u3067\u306e\u5fdc\u7b54\u306b\u306a\u3063\u3066\u3044\u307e\u3059\u3002<\/p>\n<\/div><\/div>\n\n\n\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u8a2d\u5b9a<\/div><div class=\"swell-block-step__body\">\n<p>_custom.py\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>nano .venv\/lib\/python3.10\/site-packages\/torchtune\/datasets\/_custom.py<\/code><\/pre>\n\n\n\n\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>from typing import Any, List, Mapping\nfrom torchtune.data import Message\nfrom torchtune.modules.tokenizers import Tokenizer\nfrom torchtune.datasets._chat import ChatDataset\n\ndef message_converter(\n    sample: Mapping&#091;str, Any],\n    train_on_input: bool = True,\n    ) -&gt; List&#091;Message]:\n\n    messages = &#091;\n        Message(role=\"user\", content=sample&#091;\"instruction\"]),\n        Message(role=\"assistant\", content=sample&#091;\"output\"]),\n    ]\n    return messages\n\ndef custom_dataset(\n    *,\n    tokenizer: Tokenizer,\n    source: str,\n    chat_format: str = None,\n    max_seq_len: int,\n    train_on_input: bool = True,\n    split: str,\n) -&gt; ChatDataset:\n\n    return ChatDataset(\n        tokenizer=tokenizer,\n        source=source,\n        convert_to_messages=message_converter,\n        chat_format=chat_format,\n        max_seq_len=max_seq_len,\n        train_on_input=train_on_input,\n        split=split,\n    )<\/code><\/pre>\n\n\n\n<p class=\"is-style-icon_announce\">[Ctrl + S]\u30ad\u30fc\u3067\u5909\u66f4\u5185\u5bb9\u3092\u4fdd\u5b58\u3057\u3001[Ctrl + X]\u30ad\u30fc\u3067\u7de8\u96c6\u30e2\u30fc\u30c9\u304b\u3089\u629c\u3051\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u300c__init__.py\u300d\u30d5\u30a1\u30a4\u30eb\u3092\u7de8\u96c6\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>nano .venv\/lib\/python3.10\/site-packages\/torchtune\/datasets\/__init__.py<\/code><\/pre>\n\n\n\n<p>\u300c__init__.py\u300d\u30d5\u30a1\u30a4\u30eb\u306b\u6b21\u306e\uff12\u3064\u306e\u30b3\u30fc\u30c9\u3092\u8ffd\u52a0\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<div class=\"swell-block-capbox cap_box is-style-onborder_ttl2\" data-colset=\"col2\"><div class=\"cap_box_ttl\"><span>\u8ffd\u52a0\u3059\u308b\u30b3\u30fc\u30c9<\/span><\/div><div class=\"cap_box_content\">\n<p>from torchtune.datasets._custom import custom_dataset<br>custom_dataset<\/p>\n<\/div><\/div>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>from torchtune.datasets._alpaca import alpaca_cleaned_dataset, alpaca_dataset\nfrom torchtune.datasets._chat import chat_dataset, ChatDataset\nfrom torchtune.datasets._cnn_dailymail import cnn_dailymail_articles_dataset\nfrom torchtune.datasets._concat import ConcatDataset\nfrom torchtune.datasets._grammar import grammar_dataset\nfrom torchtune.datasets._instruct import instruct_dataset, InstructDataset\nfrom torchtune.datasets._packed import PackedDataset\nfrom torchtune.datasets._preference import PreferenceDataset\nfrom torchtune.datasets._samsum import samsum_dataset\nfrom torchtune.datasets._slimorca import slimorca_dataset\nfrom torchtune.datasets._stack_exchanged_paired import stack_exchanged_paired_dataset\nfrom torchtune.datasets._text_completion import (\n    text_completion_dataset,\n    TextCompletionDataset,\n)\nfrom torchtune.datasets._wikitext import wikitext_dataset\nfrom torchtune.datasets._custom import custom_dataset <span class=\"swl-marker mark_yellow\">#\u8ffd\u52a0\n<\/span>\n__all__ = &#091;\n    \"alpaca_dataset\",\n    \"alpaca_cleaned_dataset\",\n    \"grammar_dataset\",\n    \"samsum_dataset\",\n    \"stack_exchanged_paired_dataset\",\n    \"InstructDataset\",\n    \"slimorca_dataset\",\n    \"ChatDataset\",\n    \"instruct_dataset\",\n    \"chat_dataset\",\n    \"text_completion_dataset\",\n    \"TextCompletionDataset\",\n    \"cnn_dailymail_articles_dataset\",\n    \"PackedDataset\",\n    \"ConcatDataset\",\n    \"wikitext_dataset\",\n    \"PreferenceDataset\",\n    \"custom_dataset\",  <span class=\"swl-marker mark_yellow\">#\u8ffd\u52a0<\/span>\n]<\/code><\/pre>\n\n\n\n<p class=\"is-style-icon_announce\">[Ctrl + S]\u30ad\u30fc\u3067\u5909\u66f4\u5185\u5bb9\u3092\u4fdd\u5b58\u3057\u3001[Ctrl + X]\u30ad\u30fc\u3067\u7de8\u96c6\u30e2\u30fc\u30c9\u304b\u3089\u629c\u3051\u307e\u3059\u3002<\/p>\n<\/div><\/div>\n\n\n\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u8a2d\u5b9a<\/div><div class=\"swell-block-step__body\">\n<p><span class=\"swl-format-2\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306b\u4f7f\u7528\u3059\u308bYAML\u30d5\u30a1\u30a4\u30eb\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>nano torchtune\/finetune_gozaru.yaml<\/code><\/pre>\n\n\n\n<p><span class=\"swl-format-2\">\u4ee5\u4e0b\u306e\u8a18\u8ff0\u3092\u30b3\u30d4\u30fc\u3057\u3066YAML\u30d5\u30a1\u30a4\u30eb\u306b\u8cbc\u308a\u4ed8\u3051\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code># Model Arguments\nmodel:\n  _component_: torchtune.models.llama3.qlora_llama3_8b\n  lora_attn_modules: &#091;\"q_proj\", \"k_proj\", \"v_proj\", \"o_proj\",\"gate_proj\", \"up_proj\", \"down_proj\",]\n  apply_lora_to_mlp: True\n  apply_lora_to_output: True\n  lora_rank: 8\n  lora_alpha: 16\n\n# Tokenizer\ntokenizer:\n  _component_: torchtune.models.llama3.llama3_tokenizer\n  path: \/app\/torchtune\/models\/original\/tokenizer.model\n\ncheckpointer:\n  _component_: torchtune.utils.FullModelMetaCheckpointer\n  checkpoint_dir: \/app\/torchtune\/models\/original\/\n  checkpoint_files: &#091;\n    consolidated.00.pth\n  ]\n  recipe_checkpoint: null\n  output_dir: \/app\/torchtune\/models\/\n  model_type: LLAMA3\nresume_from_checkpoint: False\n\n# Dataset and Sampler\ndataset:\n  _component_: torchtune.datasets.custom_dataset\n  source: \"bbz662bbz\/databricks-dolly-15k-ja-gozaru\"\n  max_seq_len: 2048\n  train_on_input: True\n  split: train\nseed: null\nshuffle: True\nbatch_size: 8\n\n# Optimizer and Scheduler\noptimizer:\n  _component_: torch.optim.AdamW\n  weight_decay: 0.01\n  lr: 2e-4\nlr_scheduler:\n  _component_: torchtune.modules.get_cosine_schedule_with_warmup\n  num_warmup_steps: 100\n\nloss:\n  _component_: torch.nn.CrossEntropyLoss\n\n# Training\nepochs: 2\nmax_steps_per_epoch: null\ngradient_accumulation_steps: 8\ncompile: False\n\n# Logging\noutput_dir: \/app\/torchtune\/output\/\nmetric_logger:\n  _component_: torchtune.utils.metric_logging.WandBLogger\n  project: llama3_torchtune\nlog_every_n_steps: 1\nlog_peak_memory_stats: False\n\n# Environment\ndevice: cuda\ndtype: bf16\nenable_activation_checkpointing: True\n\n# Profiler (disabled)\nprofiler:\n  _component_: torchtune.utils.profiler\n  enabled: False<\/code><\/pre>\n\n\n\n<p class=\"is-style-icon_announce\">[Ctrl + S]\u30ad\u30fc\u3067\u5909\u66f4\u5185\u5bb9\u3092\u4fdd\u5b58\u3057\u3001[Ctrl + X]\u30ad\u30fc\u3067\u7de8\u96c6\u30e2\u30fc\u30c9\u304b\u3089\u629c\u3051\u307e\u3059\u3002<\/p>\n\n\n\n<div class=\"swell-block-capbox cap_box is-style-onborder_ttl2\" data-colset=\"col3\"><div class=\"cap_box_ttl\"><span>\u30b3\u30fc\u30c9\u306e\u89e3\u8aac<\/span><\/div><div class=\"cap_box_content\">\n<p><span class=\"swl-format-2\">component: torchtune.models.llama3.qlora_llama3_8b<\/span><\/p>\n\n\n\n<p>\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u30d9\u30fc\u30b9\u306b\u306a\u308bLlama3\u306e\u30e2\u30c7\u30eb\u3092\u6307\u5b9a\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u30c7\u30eb\u306e\u7a2e\u985e\u3001\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u7a2e\u985e\u3068\u306b\u3088\u3063\u3066\u30b3\u30de\u30f3\u30c9\u304c\u7570\u306a\u308a\u307e\u3059\u306e\u3067\u3001\u8a73\u7d30\u306f<a href=\"https:\/\/pytorch.org\/torchtune\/0.1\/api_ref_models.html\">PyTorch\u306e\u516c\u5f0f\u30b5\u30a4\u30c8<\/a>\u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\"><strong>target_modules = [&#8220;q_proj&#8221;<\/strong>,&#8230;<\/span><\/p>\n\n\n\n<p>LoRA\u3092\u9069\u7528\u3059\u308b\u5bfe\u8c61\u306e\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u306e\u5c64\uff08Target modules\uff09\u3092\u6307\u5b9a\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u3059\u3079\u3066\u306e\u7dda\u5f62\u5c64\u306bLoRA\u3092\u5bfe\u8c61\u306b\u3059\u308b\u3053\u3068\u3067\u30e2\u30c7\u30eb\u306e\u9069\u5fdc\u54c1\u8cea\u304c\u5411\u4e0a\u3059\u308b\u3068\u8a00\u308f\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">r=8<\/span><\/p>\n\n\n\n<p>r\u306f\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u904e\u7a0b\u3067\u5b66\u7fd2\u3055\u308c\u308b\u4f4e\u30e9\u30f3\u30af\u884c\u5217\u306e\u30b5\u30a4\u30ba\u3092\u8868\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>r\u3092\u5927\u304d\u304f\u3059\u308b\u3068\u30e2\u30c7\u30eb\u306e\u9069\u5fdc\u54c1\u8cea\u304c\u5411\u4e0a\u3059\u308b\u50be\u5411\u304c\u3042\u308a\u307e\u3059\u304c\u3001\u5fc5\u305a\u3057\u3082\u76f4\u7dda\u7684\u306a\u95a2\u4fc2\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u3002<\/p>\n\n\n\n<p>r\u304c\u5927\u304d\u304f\u306a\u308b\u3068\u66f4\u65b0\u3055\u308c\u308b\u30d1\u30e9\u30e1\u30fc\u30bf\u304c\u5897\u3048\u308b\u305f\u3081\u3001\u30e1\u30e2\u30ea\u4f7f\u7528\u91cf\u304c\u5897\u52a0\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">lora_alpha:16<\/span><\/p>\n\n\n\n<p>LoRa\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u306eAlpha\u30d1\u30e9\u30e1\u30fc\u30bf\u306f\u3001\u5b66\u7fd2\u3057\u305f\u91cd\u307f\u3092\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p>\u591a\u304f\u306e\u6587\u732e\u3067\u306f\u3001Alpha\u3092\u8abf\u6574\u53ef\u80fd\u306a\u30d1\u30e9\u30e1\u30fc\u30bf\u3068\u3057\u3066\u6271\u3063\u3066\u304a\u3089\u305a\u3001Alpha\u309216\u306b\u56fa\u5b9a\u3057\u3066\u307e\u3059\u3002<\/p>\n\n\n\n\n\n\n\n<p><span class=\"swl-format-2\">_component_: torchtune.models.llama3.llama3_tokenizer<\/span><\/p>\n\n\n\n<p>Llama3\u306e\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u6307\u5b9a\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">source: &#8220;bbz662bbz\/databricks-dolly-15k-ja-gozaru&#8221;<\/span><\/p>\n\n\n\n<p>\u5b66\u7fd2\u306b\u4f7f\u7528\u3059\u308b\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u6307\u5b9a\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">batch_size: 8<\/span><\/p>\n\n\n\n<p>\u30d0\u30c3\u30c1\u30b5\u30a4\u30ba\u3092\u6307\u5b9a\u3057\u3066\u3044\u307e\u3059\u3002GPU\u30e1\u30e2\u30ea\u304c\u4e0d\u8db3\u3059\u308b\u5834\u5408\u306f\u3001\u5024\u3092\u5c0f\u3055\u304f\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">gradient_accumulation_steps<\/span><\/p>\n\n\n\n<p>\u52fe\u914d\u7d2f\u7a4d\u3002\u3053\u306e\u5024\u3092\u5927\u304d\u304f\u3059\u308b\u3053\u3068\u3067\u3001\u64ec\u4f3c\u7684\u306b\u30df\u30cb\u30d0\u30c3\u30c1\u306e\u30b5\u30a4\u30ba\u3092\u5927\u304d\u304f\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">component: torchtune.utils.metric_logging.WandBLogger<\/span><\/p>\n\n\n\n<p>WandB\u306b\u30ed\u30b0\u3092\u8a18\u9332\u3059\u308b\u6307\u5b9a\u3092\u3057\u3066\u3044\u307e\u3059\u3002\u4ed6\u306e\u30ed\u30ac\u30fc\u3092\u4f7f\u7528\u3059\u308b\u5834\u5408\u306e\u6307\u5b9a\u306f\u3001<a href=\"https:\/\/highreso.jp\/edgehub\/?page_id=5740\" data-type=\"page\" data-id=\"5740\">PyTorch\u516c\u5f0f\u30b5\u30a4\u30c8<\/a>\u3092\u3054\u53c2\u7167\u304f\u3060\u3055\u3044\u3002<\/p>\n<\/div><\/div>\n\n\n\n<p><span class=\"swl-format-2\">\u5b66\u7fd2\u30ed\u30b0\u306e\u7ba1\u7406\u3092\u3059\u308b\u305f\u3081\u306b\u3001WandB\u306b\u30ed\u30b0\u30a4\u30f3\u3057\u307e\u3059\u3002<\/span>\uff08WandB\u3092\u4f7f\u7528\u3057\u306a\u3044\u5834\u5408\u306f\u7701\u7565\u3057\u3066\u304f\u3060\u3055\u3044\u3002\uff09<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>wandb login ********<\/code><\/pre>\n\n\n\n<div class=\"wp-block-group is-style-big_icon_point\"><div class=\"wp-block-group__inner-container\">\n<ul class=\"wp-block-list\">\n<li>*************\u306b\u306f\u3001Wandb\u3067\u767a\u884c\u3057\u305fAPI\u30ad\u30fc\u304c\u5165\u308a\u307e\u3059\u3002<\/li>\n\n\n\n<li>\u300cwandb: Appending key for api.wandb.ai to your netrc file: \/root\/.netrc\u300d\u304c\u8fd4\u3055\u308c\u305f\u3089\u30ed\u30b0\u30a4\u30f3\u6210\u529f\u3067\u3059\u3002<\/li>\n<\/ul>\n<\/div><\/div>\n\n\n\n<p>Wandb\u3067API\u30ad\u30fc\u3092\u767a\u884c\u3059\u308b\u65b9\u6cd5\u3092\u4ee5\u4e0b\u306e\u8a18\u4e8b\u3067\u89e3\u8aac\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<div class=\"wp-block-columns\">\n<div class=\"wp-block-column\" style=\"flex-basis:100%\"><div class=\"swell-block-postLink\">\t\t\t<div class=\"p-blogCard -external\" data-type=\"type3\" data-onclick=\"clickLink\">\n\t\t\t\t<div class=\"p-blogCard__inner\">\n\t\t\t\t\t<span class=\"p-blogCard__caption\">EdgeHUB<\/span>\n\t\t\t\t\t<div class=\"p-blogCard__thumb c-postThumb\"><figure class=\"c-postThumb__figure\"><img decoding=\"async\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/wandbtsukaikata_eyecatch.jpg\" alt=\"\" class=\"c-postThumb__img u-obf-cover\" width=\"320\" height=\"180\"><\/figure><\/div>\t\t\t\t\t<div class=\"p-blogCard__body\">\n\t\t\t\t\t\t<a class=\"p-blogCard__title\" href=\"https:\/\/highreso.jp\/edgehub\/machinelearning\/wandbstartup.html\" target=\"_blank\" rel=\"noopener noreferrer\">WandB\u306e\u4f7f\u3044\u65b9\uff01\u30ed\u30b0\u30a4\u30f3\u65b9\u6cd5\u304b\u3089API\u30ad\u30fc\u767a\u884c\u3001\u30ed\u30b0\u53d6\u5f97\u307e\u3067\u7d39\u4ecb | EdgeHUB<\/a>\n\t\t\t\t\t\t<span class=\"p-blogCard__excerpt\">\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001WandB\u306e\u30a2\u30ab\u30a6\u30f3\u30c8\u4f5c\u6210\u304b\u3089\u4f7f\u3044\u65b9\u307e\u3067\u89e3\u8aac\u3057\u3066\u3044\u304d\u307e\u3059\u3002 WandB\u306f\u6a5f\u68b0\u5b66\u7fd2\u306e\u30ed\u30b0\u7ba1\u7406\u306b\u4fbf\u5229\u306a\u30c4\u30fc\u30eb\u3067\u3059\u3002 WandB\u3092\u5229\u7528\u3059\u308b\u3053\u3068\u3067\u3001\u30e2\u30c7\u30eb\u5b66\u7fd2\u306e\u30ed\u30b0\u53ce\u96c6\u3084&#8230;<\/span>\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div><\/div>\n<\/div>\n\n\n\n<p>\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>tune run lora_finetune_single_device --config torchtune\/finetune_gozaru.yaml<\/code><\/pre>\n\n\n\n<p class=\"is-style-icon_announce\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306b\u306f60\u5206\u7a0b\u5ea6\u306e\u6642\u9593\u304c\u304b\u304b\u308a\u307e\u3057\u305f\u3002<\/p>\n<\/div><\/div>\n\n\n\n<div class=\"swell-block-step__item\"><div class=\"swell-block-step__number u-bg-main\"><span class=\"__label\">STEP<\/span><\/div><div class=\"swell-block-step__title u-fz-l\">\u30e1\u30c8\u30ea\u30af\u30b9\u306e\u78ba\u8a8d<\/div><div class=\"swell-block-step__body\">\n<p>WandB\u306b\u4fdd\u5b58\u3057\u305f\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u5b9f\u884c\u4e2d\u306e\u30e1\u30c8\u30ea\u30af\u30b9\u3092\u78ba\u8a8d\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">Loss\uff08\u640d\u5931\uff09\u306f\u6700\u521d\u306e50\u30b9\u30c6\u30c3\u30d7\u3067\u6025\u901f\u306b\u6e1b\u5c11\u3057\u3001\u305d\u306e\u5f8c\u306f\u7de9\u3084\u304b\u306b\u5c0f\u3055\u304f\u306a\u308a\u53ce\u675f\u3057\u3066\u3044\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"800\" height=\"344\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune1.jpg\" alt=\"loss\" class=\"wp-image-5851\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune1.jpg 800w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune1-300x129.jpg 300w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune1-768x330.jpg 768w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/><figcaption class=\"wp-element-caption\">(\u51fa\u5178\uff1ahttps:\/\/wandb.ai\/)<\/figcaption><\/figure>\n\n\n\n<p><span class=\"swl-format-2\">GPU\u30e1\u30e2\u30ea\u306f\u3001\u5e38\u306b94%(75GB)\u3092\u4f7f\u7528\u3057\u3066\u3044\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"800\" height=\"343\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune2.jpg\" alt=\"GPU\u4f7f\u7528\u7387\" class=\"wp-image-5852\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune2.jpg 800w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune2-300x129.jpg 300w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/llama3torchtune2-768x329.jpg 768w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/><figcaption class=\"wp-element-caption\">(\u51fa\u5178\uff1ahttps:\/\/wandb.ai\/)<\/figcaption><\/figure>\n<\/div><\/div>\n<\/div>\n\n\n\n<h2 class=\"wp-block-heading\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u5f8c\u306e\u30e2\u30c7\u30eb\u306e\u78ba\u8a8d<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"600\" height=\"300\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/fall_midashi.jpg\" alt=\"\u898b\u51fa\u3057\u753b\u50cf\" class=\"wp-image-4429\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/fall_midashi.jpg 600w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/fall_midashi-300x150.jpg 300w\" sizes=\"(max-width: 600px) 100vw, 600px\" \/><\/figure>\n\n\n\n<p><span class=\"swl-format-2\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u5f8c\u306e\u30e2\u30c7\u30eb\u3067\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u3092\u3057\u3066\u3044\u304d\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<p>\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u3092\u3059\u308b\u63a8\u8ad6\u7528YAML\u30d5\u30a1\u30a4\u30eb\u3092\u4f5c\u6210\u3057\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>nano torchtune\/finetuned_inference_confg.yaml<\/code><\/pre>\n\n\n\n<p>\u4ee5\u4e0b\u306e\u30b3\u30fc\u30c9\u3092\u30b3\u30d4\u30fc\u3057\u3066\u3001YAML\u30d5\u30a1\u30a4\u30eb\u306b\u8cbc\u308a\u4ed8\u3051\u307e\u3059\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code># Model Arguments\nmodel:\n  _component_: torchtune.models.llama3.llama3_8b\n\ncheckpointer:\n  _component_: torchtune.utils.FullModelMetaCheckpointer\n  checkpoint_dir: \/app\/torchtune\/models\/\n  checkpoint_files: &#091;meta_model_0.pt] \n  recipe_checkpoint: null\n  output_dir: \/app\/torchtune\/output\/\n  model_type: LLAMA3\n\ndevice: cuda\ndtype: bf16\n\nseed: null\n\n# Tokenizer arguments\ntokenizer:\n  _component_: torchtune.models.llama3.llama3_tokenizer\n  path: \/app\/torchtune\/models\/original\/tokenizer.model\n\n# Generation arguments\nprompt: \"AirPods\u3068\u306f\u4f55\u3067\u3059\u304b\"\nmax_new_tokens: 128\ntemperature: 0.8\ntop_k: 40\nstop_tokens: \"&lt;|eot_id|&gt;\"\n\nquantizer: null<\/code><\/pre>\n\n\n\n<p><span class=\"swl-format-2\">\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u5f8c\u306e\u30e2\u30c7\u30eb\u3067\u30c6\u30ad\u30b9\u30c8\u751f\u6210\uff08\u63a8\u8ad6\uff09\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>tune run generate --config torchtune\/finetuned_inference_confg.yaml<\/code><\/pre>\n\n\n\n<div class=\"swell-block-capbox cap_box is-style-onborder_ttl2\" data-colset=\"col3\"><div class=\"cap_box_ttl\"><span>\u30b3\u30fc\u30c9\u306e\u89e3\u8aac<\/span><\/div><div class=\"cap_box_content\">\n<p><span class=\"swl-format-2\">torch.cuda.empty_cache()<\/span><\/p>\n\n\n\n<p>\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u3092\u5b9f\u884c\u3059\u308b\u524d\u306b\u3001\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3067\u4f7f\u7528\u3057\u3066\u3044\u305fGPU\u30e1\u30e2\u30ea\u3092\u30ea\u30bb\u30c3\u30c8\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<p><span class=\"swl-format-2\">FastLanguageModel.from_pretrained(&#8230;<\/span><\/p>\n\n\n\n<p>\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u5f8c\u306e\u30e2\u30c7\u30eb\u3068\u30c8\u30fc\u30af\u30ca\u30a4\u30b6\u30fc\u3092\u8aad\u307f\u8fbc\u3093\u3067\u3044\u307e\u3059\u3002<\/p>\n<\/div><\/div>\n\n\n\n<p><span class=\"swl-format-2\">&#8220;\u89b3\u8449\u690d\u7269\u306e\u52b9\u679c\u3068\u306f\uff1f&#8221;\u3068\u3044\u3046\u30d7\u30ed\u30f3\u30d7\u30c8\u3092\u5b9f\u884c\u3057\u307e\u3059\u3002<\/span><\/p>\n\n\n\n<pre class=\"wp-block-code has-black-background-color has-background\"><code>inputs = load_tokenizer(\n    &#091;\n        alpaca_prompt.format(\n        \"\u89b3\u8449\u690d\u7269\u306e\u52b9\u679c\u3068\u306f\uff1f\",\n        \"\",\n        )\n    ], return_tensors = \"pt\").to(\"cuda\")\n\noutputs = load_model.generate(\n    **inputs, \n    max_new_tokens = 128,\n    use_cache = True,\n    temperature=0.6,\n    top_p=0.9\n)\ndecoded_outputs = load_tokenizer.batch_decode(outputs,skip_special_tokens=True)\nprint(decoded_outputs&#091;0])<\/code><\/pre>\n\n\n<div class=\"swell-block-balloon\"><div class=\"c-balloon -bln-right\" data-col=\"green\"><div class=\"c-balloon__body -speaking -border-none\"><div class=\"c-balloon__text\">\n<p>### Instruction:<br>\u89b3\u8449\u690d\u7269\u306e\u52b9\u679c\u3068\u306f\uff1f<\/p>\n<span class=\"c-balloon__shapes\"><span class=\"c-balloon__before\"><\/span><span class=\"c-balloon__after\"><\/span><\/span><\/div><\/div><\/div><\/div>\n\n<div class=\"swell-block-balloon\"><div class=\"c-balloon -bln-left\" data-col=\"gray\"><div class=\"c-balloon__body -speaking -border-none\"><div class=\"c-balloon__text\">\n<p>### Response:<br>\u89b3\u8449\u690d\u7269\u306f\u3001\u7a7a\u6c17\u3092\u6e05\u6d44\u5316\u3059\u308b\u3067\u3054\u3056\u308b\u3002<\/p>\n<span class=\"c-balloon__shapes\"><span class=\"c-balloon__before\"><\/span><span class=\"c-balloon__after\"><\/span><\/span><\/div><\/div><\/div><\/div>\n\n\n<p class=\"is-style-icon_good\">\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u7279\u5fb4\u3067\u3042\u308b\u300c\u3054\u3056\u308b\u306e\u53e3\u8abf\u300d\u304c\u53cd\u6620\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u751f\u6210AI\u30fbLLM\u306e\u30b3\u30b9\u30c8\u3067\u304a\u56f0\u308a\u306a\u3089<\/h2>\n\n\n\n<p>GPU\u306e\u30b9\u30da\u30c3\u30af\u4e0d\u8db3\u3067\u751f\u6210AI\u306e\u958b\u767a\u304c\u601d\u3046\u3088\u3046\u306b\u9032\u307e\u306a\u3044\u3053\u3068\u306f\u3042\u308a\u307e\u305b\u3093\u304b\uff1f<\/p>\n\n\n\n<p>\u305d\u3093\u306a\u3068\u304d\u306b\u306f\u3001\u9ad8\u6027\u80fd\u306aGPU\u3092\u30ea\u30fc\u30ba\u30ca\u30d6\u30eb\u306a\u4fa1\u683c\u3067\u4f7f\u3048\u308bGPU\u30af\u30e9\u30a6\u30c9\u30b5\u30fc\u30d3\u30b9\u304c\u304a\u3059\u3059\u3081\u3067\u3059\uff01<\/p>\n\n\n\n<div class=\"p-blogParts post_content\" data-partsID=\"4543\">\n<div class=\"swell-block-capbox cap_box\" data-colset=\"col2\"><div class=\"cap_box_ttl\"><span>GPUSOROBAN<\/span><\/div><div class=\"cap_box_content\">\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" width=\"800\" height=\"283\" src=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/gpusorobanpr_1.jpg\" alt=\"GPUSOROBAN\" class=\"wp-image-4552\" srcset=\"https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/gpusorobanpr_1.jpg 800w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/gpusorobanpr_1-300x106.jpg 300w, https:\/\/highreso.jp\/edgehub\/wp-content\/uploads\/2024\/06\/gpusorobanpr_1-768x272.jpg 768w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/><\/figure>\n\n\n\n<p>GPUSOROBAN\u306f\u3001<span class=\"swl-format-1\"><span class=\"swl-marker mark_yellow\">\u751f\u6210AI\u30fbLLM\u5411\u3051\u306e\u9ad8\u901fGPU\u3092\u696d\u754c\u6700\u5b89\u7d1a\u306e\u6599\u91d1\u3067\u4f7f\u7528\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/span><\/span><\/p>\n\n\n\n<p>\u30a4\u30f3\u30bf\u30fc\u30cd\u30c3\u30c8\u74b0\u5883\u3055\u3048\u3042\u308c\u3070\u3001\u30af\u30e9\u30a6\u30c9\u74b0\u5883\u306eGPU\u30b5\u30fc\u30d0\u30fc\u3092\u3059\u3050\u306b\u5229\u7528\u53ef\u80fd\u3067\u3059\u3002<\/p>\n\n\n\n<p>\u5927\u898f\u6a21\u306a\u8a2d\u5099\u6295\u8cc7\u306e\u5fc5\u8981\u304c\u306a\u304f\u3001\u7169\u96d1\u306a\u30b5\u30fc\u30d0\u30fc\u7ba1\u7406\u304b\u3089\u3082\u89e3\u653e\u3055\u308c\u307e\u3059\u3002<\/p>\n\n\n\n<div class=\"swell-block-columns\"><div class=\"swell-block-columns__inner\">\n<div class=\"swell-block-column swl-has-mb--s\">\n<div class=\"swell-block-button blue_ -size-l is-style-btn_normal\"><a href=\"https:\/\/soroban.highreso.jp\/?utm_medium=edgehub&amp;utm_source=edgehub&amp;utm_campaign=highend_boxa&amp;adid=edgehub\" class=\"swell-block-button__link\"><span>\u516c\u5f0f\u30b5\u30a4\u30c8\u3092\u307f\u308b<\/span><\/a><\/div>\n<\/div>\n\n\n\n<div class=\"swell-block-column swl-has-mb--s\">\n<div class=\"swell-block-button red_ -size-l is-style-btn_normal\"><a href=\"https:\/\/soroban.highreso.jp\/download\/compute-catalog\/?utm_medium=edgehub&amp;utm_source=edgehub&amp;utm_campaign=highend_boxa&amp;adid=edgehub\" class=\"swell-block-button__link\"><span>\u8cc7\u6599\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9<\/span><\/a><\/div>\n<\/div>\n<\/div><\/div>\n<\/div><\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001PyTorch\u306eTorchtune\u3092\u4f7f\u3063\u305fLlama3\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0(QLoRA)\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002 Torchtune\u3092\u4f7f\u3063\u3066\u7c21\u5358\u306b\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u5b9f\u88c5\u304c\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002 Torchtun [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":5835,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"swell_btn_cv_data":"","footnotes":""},"categories":[20],"tags":[],"class_list":["post-5661","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-machinelearning"],"_links":{"self":[{"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/posts\/5661","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/comments?post=5661"}],"version-history":[{"count":2,"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/posts\/5661\/revisions"}],"predecessor-version":[{"id":8079,"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/posts\/5661\/revisions\/8079"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/media\/5835"}],"wp:attachment":[{"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/media?parent=5661"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/categories?post=5661"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/highreso.jp\/edgehub\/wp-json\/wp\/v2\/tags?post=5661"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}