[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-build-small-language-model-deepmind-zh":3,"article-related-build-small-language-model-deepmind-zh":29,"series-research-595ebbe1-bb81-4b4f-8f42-43a1b3820542":74},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":22,"views":26,"created_at":27,"published_at":28,"topic_cluster_id":11},"595ebbe1-bb81-4b4f-8f42-43a1b3820542","build-small-language-model-deepmind-zh","用 DeepMind 做出小型語言模型","\u003Cp data-speakable=\"summary\">這篇帶你在 \u003Ca href=\"\u002Ftag\u002Fgoogle\">Google\u003C\u002Fa> \u003Ca href=\"\u002Ftag\u002Fskills\">Skills\u003C\u002Fa> 依序完成 DeepMind 的 AI Research Foundations，建立小型語言\u003Ca href=\"\u002Fnews\u002Fanthropic-open-model-fight-lonely-ai-stance-zh\">模型的\u003C\u002Fa>基礎流程。\u003C\u002Fp>\u003Cp>這篇給已會 Python 的大學生、社群學習者與技術開發者看。照做完，你會知道要去哪裡進入 \u003Ca href=\"\u002Ftag\u002Fgoogle-deepmind\">Google DeepMind\u003C\u002Fa> 的學習路徑、三門課怎麼排、以及如何把課程概念轉成可執行的 Python 練習。\u003C\u002Fp>\u003Ch2>開始之前\u003C\u002Fh2>\u003Cul>\u003Cli>Google 帳號，可登入 \u003Ca href=\"https:\u002F\u002Fwww.skills.google\" target=\"_blank\" rel=\"noreferrer\">Google Skills\u003C\u002Fa>\u003C\u002Fli>\u003Cli>Python 3.10+\u003C\u002Fli>\u003Cli>能使用瀏覽器開啟 \u003Ca href=\"https:\u002F\u002Fwww.skills.google\u002Fcollections\u002Fdeepmind\" target=\"_blank\" rel=\"noreferrer\">DeepMind collection\u003C\u002Fa>\u003C\u002Fli>\u003Cli>具備基礎電腦科學、數學或物理課程背景\u003C\u002Fli>\u003Cli>可選：Google Cloud Console 帳號，方便延伸到雲端工作流\u003C\u002Fli>\u003C\u002Ful>\u003Cp>先確認可以登入，因為這條學習路徑會用登入狀態保存進度與課程完成度。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785522760396-cgum.png\" alt=\"用 DeepMind 做出小型語言模型\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Ch2>Step 1: 開啟 DeepMind 學習路徑\u003C\u002Fh2>\u003Cp>目的：先進到官方課程集合，確保你走的是 DeepMind 設計好的順序。\u003C\u002Fp>\u003Cp>打開 \u003Ca href=\"https:\u002F\u002Fwww.skills.google\u002Fcollections\u002Fdeepmind\" target=\"_blank\" rel=\"noreferrer\">Google Skills 的 DeepMind collection\u003C\u002Fa>，查看路徑名稱、簡介與課程清單。\u003C\u002Fp>\u003Cp>你應該看到 Google DeepMind AI Research Foundations 路徑，以及 \u003Cstrong>Start path\u003C\u002Fstrong> 入口。\u003C\u002Fp>\u003Ch2>Step 2: 啟動 AI Research Foundations\u003C\u002Fh2>\u003Cp>目的：把學習進度綁到你的帳號，讓三門課的完成狀態可追蹤。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785522761351-jqf4.png\" alt=\"用 DeepMind 做出小型語言模型\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>點選 \u003Cstrong>Start path\u003C\u002Fstrong>，若系統要求就先登入。這條路徑會帶你理解 AI 技術背後的基礎，並建立模型製作能力。\u003C\u002Fp>\u003Cp>你應該看到路徑頁面展開，第一門課已可直接進入。\u003C\u002Fp>\u003Ch2>Step 3: 依序完成三門課\u003C\u002Fh2>\u003Cp>目的：按官方建議順序學習，讓每個概念都接在前\u003Ca href=\"\u002Fnews\u002Fretoken-one-token-visual-retrieval-zh\">一個\u003C\u002Fa>基礎上。\u003C\u002Fp>\u003Cp>請依序完成這三門課：\u003Cstrong>01 Build Your Own Small Language Model\u003C\u002Fstrong>、\u003Cstrong>02 Represent Your Language Data\u003C\u002Fstrong>、\u003Cstrong>03 Design And Train Neural Networks\u003C\u002Fstrong>。\u003C\u002Fp>\u003Cp>你應該看到每門課在進度列上顯示 started 或 completed，之後再進下一門。\u003C\u002Fp>\u003Ch2>Step 4: 在 Python 重做核心概念\u003C\u002Fh2>\u003Cp>目的：把理論轉成可操作的程式練習，建立你自己的小型實驗環境。\u003C\u002Fp>\u003Cp>在你的 Python 環境裡，先用 tokenization、資料表示與簡單訓練迴圈重做課程概念。範例要小，重點放在語言模型的機制，不要先碰複雜基礎設施。\u003C\u002Fp>\u003Cpre>\u003Ccode>import torch\nimport torch.nn as nn\n\nvocab_size = 1000\nembed_dim = 128\nmodel = nn.Sequential(\n    nn.Embedding(vocab_size, embed_dim),\n    nn.Linear(embed_dim, vocab_size)\n)\nprint(model)\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>你應該看到終端機或 Notebook 印出一個最小模型定義，代表你的環境已可做小規模實驗。\u003C\u002Fp>\u003Ch2>Step 5: 檢查成果並安排下一個專案\u003C\u002Fh2>\u003Cp>目的：把完成課程\u003Ca href=\"\u002Fnews\u002Fopenai-newsroom-announcements-digest-zh\">變成\u003C\u002Fa>下一個可執行計畫，而不只是拿到徽章。\u003C\u002Fp>\u003Cp>回到 Google Skills 檢視你的學習成果，接著決定要重看課程、把概念套到 \u003Ca href=\"\u002Ftag\u002Fgoogle-cloud\">Google Cloud\u003C\u002Fa>，或把練習延伸成更完整的訓練流程。\u003C\u002Fp>\u003Cp>你應該看到已完成的路徑，以及對語言資料、神經網路與模型訓練如何串起來的更清楚理解。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>指標\u003C\u002Fth>\u003Cth>基準／優化前\u003C\u002Fth>\u003Cth>結果／優化後\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>課程順序\u003C\u002Ftd>\u003Ctd>零散自學\u003C\u002Ftd>\u003Ctd>三門 DeepMind 課程按序完成\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>學習焦點\u003C\u002Ftd>\u003Ctd>泛用 AI 興趣\u003C\u002Ftd>\u003Ctd>語言模型、資料表示、神經網路設計\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>適用對象\u003C\u002Ftd>\u003Ctd>任何學習者\u003C\u002Ftd>\u003Ctd>會 Python 的學生與技術社群學習者\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>常見錯誤\u003C\u002Fh2>\u003Cul>\u003Cli>\u003Cstrong>跳過登入。\u003C\u002Fstrong> 修法：先用 Google 帳號登入，才能保存路徑與成就。\u003C\u002Fli>\u003Cli>\u003Cstrong>先開第 03 課。\u003C\u002Fstrong> 修法：照清單順序學，因為後面的內容會預設你已懂前面的概念。\u003C\u002Fli>\u003Cli>\u003Cstrong>一開始就做大模型。\u003C\u002Fstrong> 修法：先用小型 Python 範例與簡單訓練迴圈，先看懂核心機制再擴大。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>接下來可以看什麼\u003C\u002Fh2>\u003Cp>完成這條路徑後，可以再找 Google Skills 上的 responsible AI、雲端部署與模型評估主題，把基礎延伸成可上線的工作流。\u003C\u002Fp>","這篇帶你在 Google Skills 依序完成 DeepMind 的 AI Research Foundations，建立小型語言模型的基礎流程。","www.skills.google","https:\u002F\u002Fwww.skills.google\u002Fcollections\u002Fdeepmind",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785522760396-cgum.png","research","zh","1d80b13f-dbff-4c14-b19f-fb0fdf8ea4b8",[17,18,19,20,21],"Google Skills","DeepMind","Python","language model","neural network",[23,24,25],"先登入 Google Skills，再進入 DeepMind collection，避免進度無法保存。","三門課要依序完成，才能把語言資料、表示法與訓練流程串起來。","用小型 Python 範例重做概念，最容易把課程內容轉成自己的實作能力。",0,"2026-07-31T18:32:17.616702+00:00","2026-07-31T18:32:17.607+00:00",{"tags":30,"relatedLang":33,"relatedPosts":37},[31],{"name":19,"slug":32},"python",{"id":15,"slug":34,"title":35,"language":36},"build-small-language-model-deepmind-en","Build a Small Language Model with DeepMind","en",[38,44,50,56,62,68],{"id":39,"slug":40,"title":41,"cover_image":42,"image_url":42,"created_at":43,"category":13},"27ea61f5-fd76-4df9-a9d0-05f049ca9a66","pac-man-humanoid-dodgeball-safety-zh","PAC-MAN 讓人形機器人閃球更安全","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785481375867-2a5a.png","2026-07-31T07:02:32.393678+00:00",{"id":45,"slug":46,"title":47,"cover_image":48,"image_url":48,"created_at":49,"category":13},"d0e6aae8-7bf9-42bc-a8ff-16bfd8e32c9f","retoken-one-token-visual-retrieval-zh","ReToken：一個 token 抓回視覺脈絡","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785479571709-eug9.png","2026-07-31T06:32:24.178463+00:00",{"id":51,"slug":52,"title":53,"cover_image":54,"image_url":54,"created_at":55,"category":13},"274e17d7-3802-4c6d-acfb-b09f4d6e924d","ml-trace-seiberg-dualities-zh","ML 幫忙追 Seiberg 對偶","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785477769825-coi8.png","2026-07-31T06:02:28.09076+00:00",{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":13},"59958bf8-ca7f-4d66-a170-e54eaa4c1b77","fruitfly-inspired-regression-without-heavy-models-zh","果蠅啟發回歸：用模式匹配省算力","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785394980008-nvqp.png","2026-07-30T07:02:27.028832+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":13},"f3cf3f4d-31fc-4666-8132-57c69ed66f4d","mental-world-modeling-simulating-minds-zh","世界模型不只看場景，也要看心智","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785393187520-8rd3.png","2026-07-30T06:32:29.057837+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":13},"08ceac3e-dd49-42e7-b976-e962eae021ca","pretrain-q-functions-online-rl-finetuning-zh","Q 函數不一定要先預訓練","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785391369404-txyj.png","2026-07-30T06:02:23.775888+00:00",[75,80,85,90,95,100,105,110,115,120],{"id":76,"slug":77,"title":78,"created_at":79},"f18dbadb-8c59-4723-84a4-6ad22746c77a","deepmind-bets-on-continuous-learning-ai-2026-zh","DeepMind 押注 2026 連續學習 AI","2026-03-26T08:16:02.367355+00:00",{"id":81,"slug":82,"title":83,"created_at":84},"f4a106cb-02a6-4508-8f39-9720a0a93cee","ml-papers-of-the-week-github-research-desk-zh","每週 ML 論文清單，為何紅到 GitHub","2026-03-27T01:11:39.284175+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"c4f807ca-4e5f-47f1-a48c-961cf3fc44dc","ai-ml-conferences-to-watch-in-2026-zh","2026 AI 研討會投稿時程整理","2026-03-27T01:51:53.874432+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"cf046742-efb2-4753-aef9-caed5da5e32e","adaptive-block-scaled-data-types-zh","IF4：神經網路量化的聰明選擇","2026-03-31T06:00:36.990273+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"53a0dc54-0371-4e40-8d5e-74e94a73840c","geometry-aware-similarity-metrics-for-neural-representations-zh","超越距離測量：用微分幾何重新理解神經網路","2026-03-31T06:01:01.241968+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"fee7d472-a775-4b1d-bbc2-1e8bca1bbf8b","on-the-fly-repulsion-in-the-contextual-space-for-rich-divers-zh","讓AI繪圖更有創意：用排斥力提升生成多樣性","2026-03-31T06:01:25.439673+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"a9901203-d69b-447b-8854-15d14eab32b4","vision-aided-beam-prediction-cnn-eca-zh","影像輔助波束預測升級 CNN","2026-04-01T10:00:25.8073+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"b55e7dd4-0a24-4b3d-804d-b0309a03f498","triple-band-fss-mimo-antenna-sub-6-ghz-zh","三頻 FSS MIMO 天線瞄準 sub-6 GHz","2026-04-01T13:18:36.857305+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"f68290bd-e7f3-4b30-ba22-dcd4e0130a66","openclaw-1299-repos-eight-weeks-analysis-zh","OpenClaw 1299 個 Repo 的資料解讀","2026-04-02T05:03:45.208411+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"ed9f80eb-eb02-4d35-8ad4-0ddf428751dd","beam-coherence-aware-combining-mmwave-mimo-zh","毫米波 MIMO 的雙階合併法","2026-04-02T05:27:26.897188+00:00"]