Buckets:

hf-doc-build
/

doc-dev

Files

xet

hf-doc-build/doc-dev / course /pr_1069 /th /chapter6 /5.html

rtrm

22 days ago

download

raw

97.1 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Byte-Pair Encoding tokenization","local":"byte-pair-encoding-tokenization","sections":[{"title":"อัลกอริทึมที่ใช้ในการเทรน","local":"อลกอรทมทใชในการเทรน","sections":[],"depth":2},{"title":"Tokenization algorithm","local":"tokenization-algorithm","sections":[],"depth":2},{"title":"การสร้าง BPE (Implementing BPE)","local":"การสราง-bpe-implementing-bpe","sections":[],"depth":2}],"depth":1}">
	<link href="/docs/course/pr_1069/th/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/entry/start.eeb02c13.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/scheduler.37c15a92.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/singletons.d692cc64.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/index.18351ede.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/paths.f4699845.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/entry/app.25e78aa9.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/index.2bf4358c.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/nodes/0.d0785757.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/each.e59479a4.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/nodes/40.31dc83d6.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/Tip.363c041f.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/Youtube.1e50a667.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/CodeBlock.4e987730.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/CourseFloatingBanner.6add7356.js">
	<link rel="modulepreload" href="/docs/course/pr_1069/th/_app/immutable/chunks/getInferenceSnippets.24b50994.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Byte-Pair Encoding tokenization","local":"byte-pair-encoding-tokenization","sections":[{"title":"อัลกอริทึมที่ใช้ในการเทรน","local":"อลกอรทมทใชในการเทรน","sections":[],"depth":2},{"title":"Tokenization algorithm","local":"tokenization-algorithm","sections":[],"depth":2},{"title":"การสร้าง BPE (Implementing BPE)","local":"การสราง-bpe-implementing-bpe","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="byte-pair-encoding-tokenization" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#byte-pair-encoding-tokenization"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Byte-Pair Encoding tokenization</span></h1> <div class="flex space-x-1 absolute z-10 right-0 top-0"><a href="https://discuss.huggingface.co/t/chapter-6-questions" target="_blank"><img alt="Ask a Question" class="!m-0" src="https://img.shields.io/badge/Ask%20a%20question-ffcb4c.svg?logo=data:image/svg+xml;base64,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"></a> <a href="https://colab.research.google.com/github/huggingface/notebooks/blob/master/course/th/chapter6/section5.ipynb" target="_blank"><img alt="Open In Colab" class="!m-0" src="https://colab.research.google.com/assets/colab-badge.svg"></a> <a href="https://studiolab.sagemaker.aws/import/github/huggingface/notebooks/blob/master/course/th/chapter6/section5.ipynb" target="_blank"><img alt="Open In Studio Lab" class="!m-0" src="https://studiolab.sagemaker.aws/studiolab.svg"></a></div> <p data-svelte-h="svelte-12o0prt">ดั้งเดิมแล้ว Byte-Pair Encoding (BPE) เป็นอัลกอริทึมที่ถูกสร้างเพื่อใช้บีบอัดข้อความให้เล็กลง (compress texts) ภายหลัง OpenAI ได้นำอัลกอริทึมนี้มาใช้ในการตัดคำ ในขั้นตอนเตรียมข้อมูลเพื่อเทรน GPT อัลกอริทึมตัวนี้ยังถูกนำมาใช้อย่างกว้างขวางกับโมเดลประเภท Transformer เช่น GPT, GPT-2, RoBERTa, BART, และ DeBERTa</p> <iframe class="w-full xl:w-4/6 h-80" src="https://www.youtube-nocookie.com/embed/HEikzVL-lZU" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-waucbj">💡 บทนี้จะพูดถึง BPE อย่างละเอียด เราจะเจาะลึกถึงไปถึงการ implement อัลกอริทึมนี้ คุณสามารถข้ามไปตอนท้ายได้ ถ้าคุณสนใจเพียงแค่ภาพรวมคร่าวๆเท่านั้น</p></div> <h2 class="relative group"><a id="อลกอรทมทใชในการเทรน" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#อลกอรทมทใชในการเทรน"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>อัลกอริทึมที่ใช้ในการเทรน</span></h2> <p data-svelte-h="svelte-ui9660">BPE เริ่มการเทรนด้วยการคำนวณรายการของคำที่อยู่ในคลังข้อมูล (คลังข้อมูลจะต้องผ่านการ normalization และ pre-tokenization มาแล้ว) จากนั้นมันจะเริ่มสร้างชุดคำศัพท์ (vocabulary) จากตัวอักษรที่อยู่ในแต่ละคำ มาดูตัวอย่างง่ายๆกัน เราจะสมมติว่าคลังข้อมูลของเรามีเพียงห้าคำเท่านั้น :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-string">"hug"</span>, <span class="hljs-string">"pug"</span>, <span class="hljs-string">"pun"</span>, <span class="hljs-string">"bun"</span>, <span class="hljs-string">"hugs"</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1mi54up">vocabulary ตั้งต้นสำหรับชุดข้อมูลนี้คือ <code>["b", "g", "h", "n", "p", "s", "u"]</code> ในการใช้งานจริง vocabulary ตั้งต้นจะประกอบด้วยตัวอักษร ASCII เป็นอย่างต่ำ หรืออาจจะมีตัวอักษร Unicode ได้ด้วย</p> <p data-svelte-h="svelte-568bhx">ถ้าข้อความใหม่ที่คุณต้องการจะตัดคำ มีสัญลักษณ์ที่ไม่ได้อยู่ใน training corpus สัญลักษณ์พวกนี้จะถูกแปลงเป็น unknown token นี่เป็นเหตุผลว่าทำไมโมเดล NLP จึงประมวลผลข้อความที่มีอีโมจิได้ไม่ดีนัก</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-1vpq7rj">Tokenizer ของ GPT-2 และ RoBERTa (ซึ่งค่อนข้างคล้ายกัน) มีวิธีการจัดการกับปัญหานี้ได้อย่างประสิทธิภาพ มันจะไม่มองแต่ละคำเป็น Unicode แต่จะมองว่าเป็น byte การทำแบบนี้ทำให้ vocabulary ตั้งต้น มีขนาดที่เล็ก (256) แต่ยังสามารถบันทึกทุกๆสัญลักษณ์ได้ โดยไม่ต้องแปลงสัญลักษณ์พิเศษต่างๆเป็น unknown token เทคนิคนี้เรียกว่า <em>byte-level BPE</em></p></div> <p data-svelte-h="svelte-19vck82">หลังจากสร้าง vocabulary ตั้งต้นแล้ว เราจะเพิ่ม token ใหม่ๆ เข้าไปจนว่าจะได้ vocabulary ขนาดใหญ่พอกับที่เราต้องการ โดยเราจะเทรน BPE ให้เรียน กฎที่เรียกว่า <em>merges</em> ซึ่งเป็นกฎสำหรับการรวมสองหน่วยใน vocabulary เข้าด้วยกัน
	ตอนช่วงเริ่มต้น กฎ merges พวกนี้จะสร้างคำย่อยที่ประกอบด้วยตัวอักษรสองตัว ระหว่างที่เราเทรนต่อไปเรื่อยๆ คำย่อยที่ได้ก็จะยาวขึ้น
	ในแต่ละรอบของการเทรน BPE จะคำนวณหาคู่ของคำย่อยที่พบบ่อยที่สุด (คู่ของคำย่อย ในที่นี้เราหมายถึง token ที่อยู่ติดกัน)
	คู่ที่มีความถี่มากที่สุดจะถูกรวมเข้าด้วยกัน จากนั้นโมเดลจะทำแบบเดิมอีกในการเทรนรอบต่อไป</p> <p data-svelte-h="svelte-qclq5r">กลับมาที่ตัวอย่างของเรา สมมติว่าแต่ละคำมีความถี่ดังต่อไปนี้ :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->(<span class="hljs-string">"hug"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">10</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"pug"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">5</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"pun"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">12</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"bun"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">4</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"hugs"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">5</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1hblxvc">ซึ่งแปลว่า <code>"hug"</code> พบ 10 ครั้งใน corpus, <code>"pug"</code> พบ 5 ครั้ง, <code>"pun"</code> พบ 12 ครั้ง, <code>"bun"</code> พบ 4 ครั้ง, และ <code>"hugs"</code> พบ 5 ครั้ง</p> <p data-svelte-h="svelte-14fxlqe">เราจะเริ่มการเทรน โดยแยกแต่ละคำออกเป็นตัวอักษร (ตัวอักษรจะต้องมาจาก vocabulary ตั้งต้นที่เราสร้างมาก่อนหน้านี้แล้ว) ตอนนี้คุณจะเห็นว่าแต่ละคำถูกแปลงเป็น list ที่ประกอบไปด้วยหลายๆ token</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->(<span class="hljs-string">"h"</span> <span class="hljs-string">"u"</span> <span class="hljs-string">"g"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">10</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"p"</span> <span class="hljs-string">"u"</span> <span class="hljs-string">"g"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">5</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"p"</span> <span class="hljs-string">"u"</span> <span class="hljs-string">"n"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">12</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"b"</span> <span class="hljs-string">"u"</span> <span class="hljs-string">"n"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">4</span>)<span class="hljs-punctuation">,</span> (<span class="hljs-string">"h"</span> <span class="hljs-string">"u"</span> <span class="hljs-string">"g"</span> <span class="hljs-string">"s"</span><span class="hljs-punctuation">,</span> <span class="hljs-number">5</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-bmix0l">จากนั้น เราจะดูทีละคู่ ตัวอย่างเช่น คู่ <code>("h", "u")</code> ซึ่งมาจากคำว่า <code>"hug"</code> และ <code>"hugs"</code> และพบ 15 ครั้งใน corpus
	อย่างไรก็ตาม คู่นี้ไม่ใช่คู่ที่พบบ่อยที่สุด คู่ที่พบบ่อยที่สุดคือ <code>("u", "g")</code> ซึ่งพบใน คำว่า <code>"hug"</code>, <code>"pug"</code>, และ <code>"hugs"</code> ซึ่งความถี่รวมของมันคือ 20 ครั้ง
	ดังนั้น กฎแรกของการ merge ที่ tokenizer เรียนคือ <code>("u", "g") -> "ug"</code> แปลว่ามันจะเพิ่ม <code>"ug"</code> เข้าไปใน vocabulary และใน corpus คู่นี้ก็จะถูกรวมเป็น token เดียวด้วย</p> <p data-svelte-h="svelte-7b8c15">หลังจากขั้นตอนนี้ vocabulary และ corpus จะมีค่าดังนี้ :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-symbol">Vocabulary:</span> [<span class="hljs-string">"b"</span>, <span class="hljs-string">"g"</span>, <span class="hljs-string">"h"</span>, <span class="hljs-string">"n"</span>, <span class="hljs-string">"p"</span>, <span class="hljs-string">"s"</span>, <span class="hljs-string">"u"</span>, <span class="hljs-string">"ug"</span>]
	<span class="hljs-symbol">Corpus:</span> (<span class="hljs-string">"h"</span> <span class="hljs-string">"ug"</span>, <span class="hljs-number">10</span>), (<span class="hljs-string">"p"</span> <span class="hljs-string">"ug"</span>, <span class="hljs-number">5</span>), (<span class="hljs-string">"p"</span> <span class="hljs-string">"u"</span> <span class="hljs-string">"n"</span>, <span class="hljs-number">12</span>), (<span class="hljs-string">"b"</span> <span class="hljs-string">"u"</span> <span class="hljs-string">"n"</span>, <span class="hljs-number">4</span>), (<span class="hljs-string">"h"</span> <span class="hljs-string">"ug"</span> <span class="hljs-string">"s"</span>, <span class="hljs-number">5</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1bqhc5z">ตอนนี้ จะเห็นว่าเรามีคู่ที่เมื่อรวมกันจะได้ token ที่ยาวกว่าสองตัวอักษร ตัวอย่างเช่น คู่ <code>("h", "ug")</code> ซึ่งพบ 15 ครั้งใน corpus
	อย่างไรก็ตาม คู่ที่พบบ่อยที่สุดคือ <code>("u", "n")</code> ซึ่งพบ 16 ครั้ง ดังนั้นกฎที่สองก็คือ <code>("u", "n") -> "un"</code> หลังจากที่เราเพิ่มคู่นี้ไปใน vocabulary และ merge token ใน corpus เข้าด้วยกันแล้ว เราจะได้ :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-symbol">Vocabulary:</span> [<span class="hljs-string">"b"</span>, <span class="hljs-string">"g"</span>, <span class="hljs-string">"h"</span>, <span class="hljs-string">"n"</span>, <span class="hljs-string">"p"</span>, <span class="hljs-string">"s"</span>, <span class="hljs-string">"u"</span>, <span class="hljs-string">"ug"</span>, <span class="hljs-string">"un"</span>]
	<span class="hljs-symbol">Corpus:</span> (<span class="hljs-string">"h"</span> <span class="hljs-string">"ug"</span>, <span class="hljs-number">10</span>), (<span class="hljs-string">"p"</span> <span class="hljs-string">"ug"</span>, <span class="hljs-number">5</span>), (<span class="hljs-string">"p"</span> <span class="hljs-string">"un"</span>, <span class="hljs-number">12</span>), (<span class="hljs-string">"b"</span> <span class="hljs-string">"un"</span>, <span class="hljs-number">4</span>), (<span class="hljs-string">"h"</span> <span class="hljs-string">"ug"</span> <span class="hljs-string">"s"</span>, <span class="hljs-number">5</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1g4lab1">ตอนนี้ คู่ที่พบบ่อยที่สุดคือ <code>("h", "ug")</code> ดังนั้นกฎที่ได้คือ <code>("h", "ug") -> "hug"</code> ซึ่งจะทำให้เราได้ token ที่มีสามตัวอักษร หลังจากการ merge เราจะได้ :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-symbol">Vocabulary:</span> [<span class="hljs-string">"b"</span>, <span class="hljs-string">"g"</span>, <span class="hljs-string">"h"</span>, <span class="hljs-string">"n"</span>, <span class="hljs-string">"p"</span>, <span class="hljs-string">"s"</span>, <span class="hljs-string">"u"</span>, <span class="hljs-string">"ug"</span>, <span class="hljs-string">"un"</span>, <span class="hljs-string">"hug"</span>]
	<span class="hljs-symbol">Corpus:</span> (<span class="hljs-string">"hug"</span>, <span class="hljs-number">10</span>), (<span class="hljs-string">"p"</span> <span class="hljs-string">"ug"</span>, <span class="hljs-number">5</span>), (<span class="hljs-string">"p"</span> <span class="hljs-string">"un"</span>, <span class="hljs-number">12</span>), (<span class="hljs-string">"b"</span> <span class="hljs-string">"un"</span>, <span class="hljs-number">4</span>), (<span class="hljs-string">"hug"</span> <span class="hljs-string">"s"</span>, <span class="hljs-number">5</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-rt06o2">เราจะทำแบบนี้ต่อไปเรื่อยๆ จนกว่าจะได้ขนาดของ vocabulary ที่ต้องการ</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-ms4toc">✏️ <strong>ตาคุณแล้ว!</strong> คุณคิดว่ากฎ merge ต่อไปคืออะไร</p></div> <h2 class="relative group"><a id="tokenization-algorithm" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#tokenization-algorithm"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Tokenization algorithm</span></h2> <p data-svelte-h="svelte-38g5f4">การ tokenization เป็นขั้นตอนหลังจากการเทรน โดย input ใหม่จะถูก tokenize ด้วยขั้นตอนดังต่อไปนี้</p> <ol data-svelte-h="svelte-1dn790a"><li>Normalization (การปรับข้อความให้เป็นมาตรฐาน)</li> <li>Pre-tokenization (การเตรียมข้อความให้พร้อมสำหรับการ tokenize จริง)</li> <li>แยกคำออกเป็นตัวอักษรเดี่ยว</li> <li>ใช้กฎ merge ที่ได้จากการเทรนเพื่อรวมตัวอักษรที่เราได้จากขั้นตอนก่อนหน้า</li></ol> <p data-svelte-h="svelte-1c2vghc">มาดูกฎสามตัวที่เราได้จากการเทรนก่อนหน้านี้ :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-function"><span class="hljs-params">(<span class="hljs-string">"u"</span>, <span class="hljs-string">"g"</span>)</span> -></span> <span class="hljs-string">"ug"</span>
	<span class="hljs-function"><span class="hljs-params">(<span class="hljs-string">"u"</span>, <span class="hljs-string">"n"</span>)</span> -></span> <span class="hljs-string">"un"</span>
	<span class="hljs-function"><span class="hljs-params">(<span class="hljs-string">"h"</span>, <span class="hljs-string">"ug"</span>)</span> -></span> <span class="hljs-string">"hug"</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-n933xw">คำว่า<code>"bug"</code> จะถูกแยกเป็น <code>["b", "ug"]</code> ส่วนคำว่า <code>"mug"</code> จะถูกแยกเป็น <code>["[UNK]", "ug"]</code> เพราะว่า <code>"m"</code> ไม่ได้อยู่ใน vocabulary ของเรา
	้เช่นเดียวกัน คำว่า <code>"thug"</code> จะถูกแยกเป็น <code>["[UNK]", "hug"]</code> เพราะว่า <code>"t"</code> ไม่ได้อยู่ใน vocabulary กฎแรกจะรวม <code>"u"</code> และ <code>"g"</code> เข้าด้วยกัน จากนั้น <code>"hu"</code> และ <code>"g"</code> ก็จะถูกรวมเข้าด้วยกัน</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-13f5w1r">✏️ <strong>ตาคุณแล้ว!</strong> คุณคิดว่าคำว่า <code>"unhug"</code> จะถูกแยกอย่างไร</p></div> <h2 class="relative group"><a id="การสราง-bpe-implementing-bpe" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#การสราง-bpe-implementing-bpe"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>การสร้าง BPE (Implementing BPE)</span></h2> <p data-svelte-h="svelte-1e93cpt">ตอนนี้เราจะมาดูกันว่า คุณจะสามารถ implement อัลกอริทึม BPE ได้อย่างไร สิ่งที่เราจะเรียนต่อไปนี้ไม่ใช่ implementation ที่ดีที่สุด เราเพียงต้องการให้คุณเข้าใจโค้ดและเข้าใจว่า BPE ทำงานอย่างไร</p> <p data-svelte-h="svelte-5w2uhv">อันดับแรก เราต้องการ corpus ดังนั้น เราจะสร้าง corpus แบบง่ายๆขึ้นมา โดยประกอบไปด้วยไม่กี่ประโยค :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->corpus = [
	<span class="hljs-string">"This is the Hugging Face course."</span>,
	<span class="hljs-string">"This chapter is about tokenization."</span>,
	<span class="hljs-string">"This section shows several tokenizer algorithms."</span>,
	<span class="hljs-string">"Hopefully, you will be able to understand how they are trained and generate tokens."</span>,
	]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1p977ot">จากนั้นเราจะทำการ pre-tokenize corpus นี้ เพื่อแยกข้อความออกเป็นคำๆ เนื่องจากเราจะสร้าง BPE tokenizer ตามตัวที่ใช้ใน GPT-2 เราจึงต้องใช้ <code>gpt2</code> tokenizer ในการ pre-tokenize</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer

	tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"gpt2"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-97lq7n">จากนั้นเราจะคำนวณความถี่ของแต่ละคำ:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> collections <span class="hljs-keyword">import</span> defaultdict

	word_freqs = defaultdict(<span class="hljs-built_in">int</span>)

	<span class="hljs-keyword">for</span> text <span class="hljs-keyword">in</span> corpus:
	words_with_offsets = tokenizer.backend_tokenizer.pre_tokenizer.pre_tokenize_str(text)
	new_words = [word <span class="hljs-keyword">for</span> word, offset <span class="hljs-keyword">in</span> words_with_offsets]
	<span class="hljs-keyword">for</span> word <span class="hljs-keyword">in</span> new_words:
	word_freqs[word] += <span class="hljs-number">1</span>

	<span class="hljs-built_in">print</span>(word_freqs)<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->defaultdict(<span class="hljs-built_in">int</span>, {<span class="hljs-string">'This'</span>: <span class="hljs-number">3</span>, <span class="hljs-string">'Ġis'</span>: <span class="hljs-number">2</span>, <span class="hljs-string">'Ġthe'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'ĠHugging'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'ĠFace'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'ĠCourse'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'.'</span>: <span class="hljs-number">4</span>, <span class="hljs-string">'Ġchapter'</span>: <span class="hljs-number">1</span>,
	<span class="hljs-string">'Ġabout'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġtokenization'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġsection'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġshows'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġseveral'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġtokenizer'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġalgorithms'</span>: <span class="hljs-number">1</span>,
	<span class="hljs-string">'Hopefully'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">','</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġyou'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġwill'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġbe'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġable'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġto'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġunderstand'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġhow'</span>: <span class="hljs-number">1</span>,
	<span class="hljs-string">'Ġthey'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġare'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġtrained'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġand'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġgenerate'</span>: <span class="hljs-number">1</span>, <span class="hljs-string">'Ġtokens'</span>: <span class="hljs-number">1</span>})<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1ki81xp">ขั้นตอนต่อไป คือการคำนวณ vocabulary ตั้งต้น ซึ่งสร้างจากแต่ละตัวอักษรใน corpus :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->alphabet = []

	<span class="hljs-keyword">for</span> word <span class="hljs-keyword">in</span> word_freqs.keys():
	<span class="hljs-keyword">for</span> letter <span class="hljs-keyword">in</span> word:
	<span class="hljs-keyword">if</span> letter <span class="hljs-keyword">not</span> <span class="hljs-keyword">in</span> alphabet:
	alphabet.append(letter)
	alphabet.sort()

	<span class="hljs-built_in">print</span>(alphabet)<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->[ <span class="hljs-string">','</span>, <span class="hljs-string">'.'</span>, <span class="hljs-string">'C'</span>, <span class="hljs-string">'F'</span>, <span class="hljs-string">'H'</span>, <span class="hljs-string">'T'</span>, <span class="hljs-string">'a'</span>, <span class="hljs-string">'b'</span>, <span class="hljs-string">'c'</span>, <span class="hljs-string">'d'</span>, <span class="hljs-string">'e'</span>, <span class="hljs-string">'f'</span>, <span class="hljs-string">'g'</span>, <span class="hljs-string">'h'</span>, <span class="hljs-string">'i'</span>, <span class="hljs-string">'k'</span>, <span class="hljs-string">'l'</span>, <span class="hljs-string">'m'</span>, <span class="hljs-string">'n'</span>, <span class="hljs-string">'o'</span>, <span class="hljs-string">'p'</span>, <span class="hljs-string">'r'</span>, <span class="hljs-string">'s'</span>,
	<span class="hljs-string">'t'</span>, <span class="hljs-string">'u'</span>, <span class="hljs-string">'v'</span>, <span class="hljs-string">'w'</span>, <span class="hljs-string">'y'</span>, <span class="hljs-string">'z'</span>, <span class="hljs-string">'Ġ'</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1ivy2j">เราจะเพิ่ม token พิเศษเข้าไปในข้างหน้า list นี้ด้วย GPT-2 ใช้ token พิเศษคือ <code>"<\|endoftext\|>"</code> :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->vocab = [<span class="hljs-string">"<\|endoftext\|>"</span>] + alphabet.copy()<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1d6a2sn">จากนั้นเราจะแยกแต่ละคำใน corpus ให้เป็นตัวอักษร เพื่อที่เราจะได้เริ่มการเทรน :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->splits = {word: [c <span class="hljs-keyword">for</span> c <span class="hljs-keyword">in</span> word] <span class="hljs-keyword">for</span> word <span class="hljs-keyword">in</span> word_freqs.keys()}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1n9k9s1">ตอนนี้เราก็พร้อมที่จะเทรนแล้ว เราจะเริ่มด้วยการเขียนฟังก์ชันที่คำนวณความถี่ของแต่ละคู่ token :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">compute_pair_freqs</span>(<span class="hljs-params">splits</span>):
	pair_freqs = defaultdict(<span class="hljs-built_in">int</span>)
	<span class="hljs-keyword">for</span> word, freq <span class="hljs-keyword">in</span> word_freqs.items():
	split = splits[word]
	<span class="hljs-keyword">if</span> <span class="hljs-built_in">len</span>(split) == <span class="hljs-number">1</span>:
	<span class="hljs-keyword">continue</span>
	<span class="hljs-keyword">for</span> i <span class="hljs-keyword">in</span> <span class="hljs-built_in">range</span>(<span class="hljs-built_in">len</span>(split) - <span class="hljs-number">1</span>):
	pair = (split[i], split[i + <span class="hljs-number">1</span>])
	pair_freqs[pair] += freq
	<span class="hljs-keyword">return</span> pair_freqs<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1jpkvry">มาดูส่วนผลลัพธ์ (ซึ่งเป็น dictionary) กัน :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->pair_freqs = compute_pair_freqs(splits)

	<span class="hljs-keyword">for</span> i, key <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(pair_freqs.keys()):
	<span class="hljs-built_in">print</span>(<span class="hljs-string">f"<span class="hljs-subst">{key}</span>: <span class="hljs-subst">{pair_freqs[key]}</span>"</span>)
	<span class="hljs-keyword">if</span> i >= <span class="hljs-number">5</span>:
	<span class="hljs-keyword">break</span><!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->(<span class="hljs-string">'T'</span>, <span class="hljs-string">'h'</span>): <span class="hljs-number">3</span>
	(<span class="hljs-string">'h'</span>, <span class="hljs-string">'i'</span>): <span class="hljs-number">3</span>
	(<span class="hljs-string">'i'</span>, <span class="hljs-string">'s'</span>): <span class="hljs-number">5</span>
	(<span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'i'</span>): <span class="hljs-number">2</span>
	(<span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'t'</span>): <span class="hljs-number">7</span>
	(<span class="hljs-string">'t'</span>, <span class="hljs-string">'h'</span>): <span class="hljs-number">3</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ytfm66">จากนั้นเราจะหาคู่ที่พบบ่อยที่สุด ซึ่งทำได้ง่ายๆดังนี้ :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->best_pair = <span class="hljs-string">""</span>
	max_freq = <span class="hljs-literal">None</span>

	<span class="hljs-keyword">for</span> pair, freq <span class="hljs-keyword">in</span> pair_freqs.items():
	<span class="hljs-keyword">if</span> max_freq <span class="hljs-keyword">is</span> <span class="hljs-literal">None</span> <span class="hljs-keyword">or</span> max_freq < freq:
	best_pair = pair
	max_freq = freq

	<span class="hljs-built_in">print</span>(best_pair, max_freq)<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->(<span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'t'</span>) <span class="hljs-number">7</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-knui6i">ดังนั้น กฎแรกที่เราได้ก็คือ <code>('Ġ', 't') -> 'Ġt'</code> และเราจะต้องเพิ่ม <code>'Ġt'</code> เข้าไปใน vocabulary :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->merges = {(<span class="hljs-string">"Ġ"</span>, <span class="hljs-string">"t"</span>): <span class="hljs-string">"Ġt"</span>}
	vocab.append(<span class="hljs-string">"Ġt"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1gz885v">จากนั้น เราจะต้องทำการ merge คำย่อยที่อยู่ใน dictionary <code>splits</code> ด้วย โดยเราจะเขียนฟังก์ชันต่อไปนี้ :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">merge_pair</span>(<span class="hljs-params">a, b, splits</span>):
	<span class="hljs-keyword">for</span> word <span class="hljs-keyword">in</span> word_freqs:
	split = splits[word]
	<span class="hljs-keyword">if</span> <span class="hljs-built_in">len</span>(split) == <span class="hljs-number">1</span>:
	<span class="hljs-keyword">continue</span>

	i = <span class="hljs-number">0</span>
	<span class="hljs-keyword">while</span> i < <span class="hljs-built_in">len</span>(split) - <span class="hljs-number">1</span>:
	<span class="hljs-keyword">if</span> split[i] == a <span class="hljs-keyword">and</span> split[i + <span class="hljs-number">1</span>] == b:
	split = split[:i] + [a + b] + split[i + <span class="hljs-number">2</span> :]
	<span class="hljs-keyword">else</span>:
	i += <span class="hljs-number">1</span>
	splits[word] = split
	<span class="hljs-keyword">return</span> splits<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-k9g317">และนี่ก็คือผลลัพธ์จากการ merge ครั้งแรก :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->splits = merge_pair(<span class="hljs-string">"Ġ"</span>, <span class="hljs-string">"t"</span>, splits)
	<span class="hljs-built_in">print</span>(splits[<span class="hljs-string">"Ġtrained"</span>])<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->[<span class="hljs-string">'Ġt'</span>, <span class="hljs-string">'r'</span>, <span class="hljs-string">'a'</span>, <span class="hljs-string">'i'</span>, <span class="hljs-string">'n'</span>, <span class="hljs-string">'e'</span>, <span class="hljs-string">'d'</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ch2nw4">ตอนนี้เราก็มีทุกอย่างพร้อมสำหรับการเทรนแล้ว เราจะเทรนจนกว่าขนาดของ vocabulary จะเท่ากับ 50 :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->vocab_size = <span class="hljs-number">50</span>

	<span class="hljs-keyword">while</span> <span class="hljs-built_in">len</span>(vocab) < vocab_size:
	pair_freqs = compute_pair_freqs(splits)
	best_pair = <span class="hljs-string">""</span>
	max_freq = <span class="hljs-literal">None</span>
	<span class="hljs-keyword">for</span> pair, freq <span class="hljs-keyword">in</span> pair_freqs.items():
	<span class="hljs-keyword">if</span> max_freq <span class="hljs-keyword">is</span> <span class="hljs-literal">None</span> <span class="hljs-keyword">or</span> max_freq < freq:
	best_pair = pair
	max_freq = freq
	splits = merge_pair(*best_pair, splits)
	merges[best_pair] = best_pair[<span class="hljs-number">0</span>] + best_pair[<span class="hljs-number">1</span>]
	vocab.append(best_pair[<span class="hljs-number">0</span>] + best_pair[<span class="hljs-number">1</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1047qda">ผลลัพธ์ที่ได้คือ tokenizer ของเราได้เรียน 19 กฎ (vocabulary ตั้งต้นมี 31 token ซึ่งมาจากตัวอักษรที่เรามี 30 ตัวและ token พิเศษอีกหนึ่งตัว) :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-built_in">print</span>(merges)<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->{(<span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'t'</span>): <span class="hljs-string">'Ġt'</span>, (<span class="hljs-string">'i'</span>, <span class="hljs-string">'s'</span>): <span class="hljs-string">'is'</span>, (<span class="hljs-string">'e'</span>, <span class="hljs-string">'r'</span>): <span class="hljs-string">'er'</span>, (<span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'a'</span>): <span class="hljs-string">'Ġa'</span>, (<span class="hljs-string">'Ġt'</span>, <span class="hljs-string">'o'</span>): <span class="hljs-string">'Ġto'</span>, (<span class="hljs-string">'e'</span>, <span class="hljs-string">'n'</span>): <span class="hljs-string">'en'</span>,
	(<span class="hljs-string">'T'</span>, <span class="hljs-string">'h'</span>): <span class="hljs-string">'Th'</span>, (<span class="hljs-string">'Th'</span>, <span class="hljs-string">'is'</span>): <span class="hljs-string">'This'</span>, (<span class="hljs-string">'o'</span>, <span class="hljs-string">'u'</span>): <span class="hljs-string">'ou'</span>, (<span class="hljs-string">'s'</span>, <span class="hljs-string">'e'</span>): <span class="hljs-string">'se'</span>, (<span class="hljs-string">'Ġto'</span>, <span class="hljs-string">'k'</span>): <span class="hljs-string">'Ġtok'</span>,
	(<span class="hljs-string">'Ġtok'</span>, <span class="hljs-string">'en'</span>): <span class="hljs-string">'Ġtoken'</span>, (<span class="hljs-string">'n'</span>, <span class="hljs-string">'d'</span>): <span class="hljs-string">'nd'</span>, (<span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'is'</span>): <span class="hljs-string">'Ġis'</span>, (<span class="hljs-string">'Ġt'</span>, <span class="hljs-string">'h'</span>): <span class="hljs-string">'Ġth'</span>, (<span class="hljs-string">'Ġth'</span>, <span class="hljs-string">'e'</span>): <span class="hljs-string">'Ġthe'</span>,
	(<span class="hljs-string">'i'</span>, <span class="hljs-string">'n'</span>): <span class="hljs-string">'in'</span>, (<span class="hljs-string">'Ġa'</span>, <span class="hljs-string">'b'</span>): <span class="hljs-string">'Ġab'</span>, (<span class="hljs-string">'Ġtoken'</span>, <span class="hljs-string">'i'</span>): <span class="hljs-string">'Ġtokeni'</span>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-19ymfzy">ส่วน vocabulary ที่ได้จะประกอบไปด้วย token พิเศษ, ตัวอักษรตั้งต้น, และผลลัพธ์จากการ merge แต่ละครั้ง :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-built_in">print</span>(vocab)<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->[<span class="hljs-string">'<\|endoftext\|>'</span>, <span class="hljs-string">','</span>, <span class="hljs-string">'.'</span>, <span class="hljs-string">'C'</span>, <span class="hljs-string">'F'</span>, <span class="hljs-string">'H'</span>, <span class="hljs-string">'T'</span>, <span class="hljs-string">'a'</span>, <span class="hljs-string">'b'</span>, <span class="hljs-string">'c'</span>, <span class="hljs-string">'d'</span>, <span class="hljs-string">'e'</span>, <span class="hljs-string">'f'</span>, <span class="hljs-string">'g'</span>, <span class="hljs-string">'h'</span>, <span class="hljs-string">'i'</span>, <span class="hljs-string">'k'</span>, <span class="hljs-string">'l'</span>, <span class="hljs-string">'m'</span>, <span class="hljs-string">'n'</span>, <span class="hljs-string">'o'</span>,
	<span class="hljs-string">'p'</span>, <span class="hljs-string">'r'</span>, <span class="hljs-string">'s'</span>, <span class="hljs-string">'t'</span>, <span class="hljs-string">'u'</span>, <span class="hljs-string">'v'</span>, <span class="hljs-string">'w'</span>, <span class="hljs-string">'y'</span>, <span class="hljs-string">'z'</span>, <span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'Ġt'</span>, <span class="hljs-string">'is'</span>, <span class="hljs-string">'er'</span>, <span class="hljs-string">'Ġa'</span>, <span class="hljs-string">'Ġto'</span>, <span class="hljs-string">'en'</span>, <span class="hljs-string">'Th'</span>, <span class="hljs-string">'This'</span>, <span class="hljs-string">'ou'</span>, <span class="hljs-string">'se'</span>,
	<span class="hljs-string">'Ġtok'</span>, <span class="hljs-string">'Ġtoken'</span>, <span class="hljs-string">'nd'</span>, <span class="hljs-string">'Ġis'</span>, <span class="hljs-string">'Ġth'</span>, <span class="hljs-string">'Ġthe'</span>, <span class="hljs-string">'in'</span>, <span class="hljs-string">'Ġab'</span>, <span class="hljs-string">'Ġtokeni'</span>]<!-- HTML_TAG_END --></pre></div> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-7ebhk9">💡 ถ้าคุณใช้ <code>train_new_from_iterator()</code> กับ corpus เดียวกันนี้ คุณจะไม่ได้ vocabulary เดียวกัน เพราะว่าอาจจะมีหลายคู่ token ที่มีความถี่สูงสุดเท่ากัน ในตัวอย่างของเรา เราเลือกคู่แรกที่โค้ดของเราอ่านเจอ ส่วน 🤗 Tokenizers library เลือกคู่แรกโดยเรียงตาม ID</p></div> <p data-svelte-h="svelte-1g50xxw">หากเราต้องการ tokenize ข้อความใดข้อความหนึ่ง สิ่งที่ต้องทำคือ pre-tokenize จากนั้นจึงทำการ tokenize และสุดท้าย apply กฎ merge :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">tokenize</span>(<span class="hljs-params">text</span>):
	pre_tokenize_result = tokenizer._tokenizer.pre_tokenizer.pre_tokenize_str(text)
	pre_tokenized_text = [word <span class="hljs-keyword">for</span> word, offset <span class="hljs-keyword">in</span> pre_tokenize_result]
	splits = [[l <span class="hljs-keyword">for</span> l <span class="hljs-keyword">in</span> word] <span class="hljs-keyword">for</span> word <span class="hljs-keyword">in</span> pre_tokenized_text]
	<span class="hljs-keyword">for</span> pair, merge <span class="hljs-keyword">in</span> merges.items():
	<span class="hljs-keyword">for</span> idx, split <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(splits):
	i = <span class="hljs-number">0</span>
	<span class="hljs-keyword">while</span> i < <span class="hljs-built_in">len</span>(split) - <span class="hljs-number">1</span>:
	<span class="hljs-keyword">if</span> split[i] == pair[<span class="hljs-number">0</span>] <span class="hljs-keyword">and</span> split[i + <span class="hljs-number">1</span>] == pair[<span class="hljs-number">1</span>]:
	split = split[:i] + [merge] + split[i + <span class="hljs-number">2</span> :]
	<span class="hljs-keyword">else</span>:
	i += <span class="hljs-number">1</span>
	splits[idx] = split

	<span class="hljs-keyword">return</span> <span class="hljs-built_in">sum</span>(splits, [])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1xxndsq">คุณสามารถทดลองโค้ดนี้ได้กับข้อความทุกข้อความ ที่ประกอบไปด้วยตัวอักษรเท่านั้น :</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->tokenize(<span class="hljs-string">"This is not a token."</span>)<!-- HTML_TAG_END --></pre></div> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->[<span class="hljs-string">'This'</span>, <span class="hljs-string">'Ġis'</span>, <span class="hljs-string">'Ġ'</span>, <span class="hljs-string">'n'</span>, <span class="hljs-string">'o'</span>, <span class="hljs-string">'t'</span>, <span class="hljs-string">'Ġa'</span>, <span class="hljs-string">'Ġtoken'</span>, <span class="hljs-string">'.'</span>]<!-- HTML_TAG_END --></pre></div> <div class="course-tip course-tip-orange bg-gradient-to-br dark:bg-gradient-to-r before:border-orange-500 dark:before:border-orange-800 from-orange-50 dark:from-gray-900 to-white dark:to-gray-950 border border-orange-50 text-orange-700 dark:text-gray-400"><p data-svelte-h="svelte-qgvtsz">⚠️ การ implementation ในตัวอย่างของเราจะ return error ถ้าโปรแกรมอ่านเจอตัวอักษรที่ไม่มีใน vocabulary นั่นเพราะว่าเราไม่ได้เขียนโค้ดเพื่อจัดการกับกรณีแบบนี้
	ใน GPT-2 ปกติจะไม่มี unknown token แบบนี้ เพราะว่า ถ้าเราใช้ byte-level BPE เราจะไม่มีทางได้ตัวอักษรที่ unknown อย่างไรก็ตามในตัวอย่างของเรา เราไม่ได้ใช้ทุกๆ byte เพื่อสร้าง vocabulary ตั้งต้น
	อย่างไรก็ตาม หัวข้อนี้นั้นค่อนข้างลึก เราจึงจะไม่ขอพูดถึงรายละเอียดไปมากกว่านี้</p></div> <p data-svelte-h="svelte-zewf2t">นี่ก็คือ อัลกอริทึม BPE ในบทต่อไป เราจะมาดู WordPiece กัน</p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/course/blob/main/chapters/th/chapter6/5.mdx" target="_blank"><span data-svelte-h="svelte-1kd6by1"><</span> <span data-svelte-h="svelte-x0xyl0">></span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a> <p></p>

	<script>
	{
	__sveltekit_1q1y86z = {
	assets: "/docs/course/pr_1069/th",
	base: "/docs/course/pr_1069/th",
	env: {}
	};

	const element = document.currentScript.parentElement;

	const data = [null,null];

	Promise.all([
	import("/docs/course/pr_1069/th/_app/immutable/entry/start.eeb02c13.js"),
	import("/docs/course/pr_1069/th/_app/immutable/entry/app.25e78aa9.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 40],
	data,
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 97.1 kB
Xet hash:: 8446466d9fa279bb84ede00edbb6b863fc0fcdf21c437c073213554377713ce0

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.