Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Preprocess","local":"preprocess","sections":[{"title":"NLP","local":"nlp","sections":[{"title":"Tokenize","local":"tokenize","sections":[],"depth":3},{"title":"Pad","local":"pad","sections":[],"depth":3},{"title":"Truncation","local":"truncation","sections":[],"depth":3},{"title":"Costruire i tensori","local":"costruire-i-tensori","sections":[],"depth":3}],"depth":2},{"title":"Audio","local":"audio","sections":[{"title":"Ricampionamento","local":"ricampionamento","sections":[],"depth":3},{"title":"Feature extractor","local":"feature-extractor","sections":[],"depth":3},{"title":"Pad e truncate","local":"pad-e-truncate","sections":[],"depth":3}],"depth":2},{"title":"Vision","local":"vision","sections":[{"title":"Feature extractor","local":"feature-extractor","sections":[],"depth":3},{"title":"Data augmentation","local":"data-augmentation","sections":[],"depth":3}],"depth":2},{"title":"Multimodal","local":"multimodal","sections":[{"title":"Processor","local":"processor","sections":[],"depth":3}],"depth":2}],"depth":1}"> | |
| <link href="/docs/transformers/pr_33913/it/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/entry/start.cef8f86b.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/scheduler.36a0863c.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/singletons.39cd3b38.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/index.733708bb.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/paths.64312358.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/entry/app.d399d34f.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/index.9c13489a.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/nodes/0.7859fb6f.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/nodes/28.db7b5ab4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/Tip.3b06990e.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/Youtube.347c76e5.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/CodeBlock.05d8ec32.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/IconTensorflow.e7b84ee9.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/DocNotebookDropdown.653c9eec.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/globals.7f7f1b26.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/EditOnGithub.e88f2b7b.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Preprocess","local":"preprocess","sections":[{"title":"NLP","local":"nlp","sections":[{"title":"Tokenize","local":"tokenize","sections":[],"depth":3},{"title":"Pad","local":"pad","sections":[],"depth":3},{"title":"Truncation","local":"truncation","sections":[],"depth":3},{"title":"Costruire i tensori","local":"costruire-i-tensori","sections":[],"depth":3}],"depth":2},{"title":"Audio","local":"audio","sections":[{"title":"Ricampionamento","local":"ricampionamento","sections":[],"depth":3},{"title":"Feature extractor","local":"feature-extractor","sections":[],"depth":3},{"title":"Pad e truncate","local":"pad-e-truncate","sections":[],"depth":3}],"depth":2},{"title":"Vision","local":"vision","sections":[{"title":"Feature extractor","local":"feature-extractor","sections":[],"depth":3},{"title":"Data augmentation","local":"data-augmentation","sections":[],"depth":3}],"depth":2},{"title":"Multimodal","local":"multimodal","sections":[{"title":"Processor","local":"processor","sections":[],"depth":3}],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="preprocess" 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="#preprocess"><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>Preprocess</span></h1> <div class="flex space-x-1 absolute z-10 right-0 top-0"> <div class="relative colab-dropdown "> <button class=" " type="button"> <img alt="Open In Colab" class="!m-0" src="https://colab.research.google.com/assets/colab-badge.svg"> </button> </div> <div class="relative colab-dropdown "> <button class=" " type="button"> <img alt="Open In Studio Lab" class="!m-0" src="https://studiolab.sagemaker.aws/studiolab.svg"> </button> </div></div> <p data-svelte-h="svelte-wbj44x">Prima di poter usare i dati in un modello, bisogna processarli in un formato accettabile per quest’ultimo. Un modello non comprende il testo grezzo, le immagini o l’audio. Bisogna convertire questi input in numeri e assemblarli all’interno di tensori. In questa esercitazione, tu potrai:</p> <ul data-svelte-h="svelte-1k271m6"><li>Preprocessare dati testuali con un tokenizer.</li> <li>Preprocessare immagini o dati audio con un estrattore di caratteristiche.</li> <li>Preprocessare dati per attività multimodali mediante un processore.</li></ul> <h2 class="relative group"><a id="nlp" 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="#nlp"><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>NLP</span></h2> <iframe class="w-full xl:w-4/6 h-80" src="https://www.youtube-nocookie.com/embed/Yffk5aydLzg" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe> <p data-svelte-h="svelte-1lqvr4q">Lo strumento principale per processare dati testuali è un <a href="main_classes/tokenizer">tokenizer</a>. Un tokenizer inizia separando il testo in <em>tokens</em> secondo una serie di regole. I tokens sono convertiti in numeri, questi vengono utilizzati per costruire i tensori di input del modello. Anche altri input addizionali se richiesti dal modello vengono aggiunti dal tokenizer.</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-mhp41y">Se stai pensando si utilizzare un modello preaddestrato, è importante utilizzare il tokenizer preaddestrato associato. Questo assicura che il testo sia separato allo stesso modo che nel corpus usato per l’addestramento, e venga usata la stessa mappatura tokens-to-index (solitamente indicato come il <em>vocabolario</em>) come nel preaddestramento.</p></div> <p data-svelte-h="svelte-1fr955r">Iniziamo subito caricando un tokenizer preaddestrato con la classe <code>AutoTokenizer</code>. Questo scarica il <em>vocabolario</em> usato quando il modello è stato preaddestrato.</p> <h3 class="relative group"><a id="tokenize" 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="#tokenize"><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>Tokenize</span></h3> <p data-svelte-h="svelte-1fuig6i">Carica un tokenizer preaddestrato con <code>AutoTokenizer.from_pretrained()</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 --><span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| <span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"google-bert/bert-base-cased"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-11flzdh">Poi inserisci le tue frasi nel tokenizer:</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-meta">>>> </span>encoded_input = tokenizer(<span class="hljs-string">"Do not meddle in the affairs of wizards, for they are subtle and quick to anger."</span>) | |
| <span class="hljs-meta">>>> </span><span class="hljs-built_in">print</span>(encoded_input) | |
| {<span class="hljs-string">'input_ids'</span>: [<span class="hljs-number">101</span>, <span class="hljs-number">2079</span>, <span class="hljs-number">2025</span>, <span class="hljs-number">19960</span>, <span class="hljs-number">10362</span>, <span class="hljs-number">1999</span>, <span class="hljs-number">1996</span>, <span class="hljs-number">3821</span>, <span class="hljs-number">1997</span>, <span class="hljs-number">16657</span>, <span class="hljs-number">1010</span>, <span class="hljs-number">2005</span>, <span class="hljs-number">2027</span>, <span class="hljs-number">2024</span>, <span class="hljs-number">11259</span>, <span class="hljs-number">1998</span>, <span class="hljs-number">4248</span>, <span class="hljs-number">2000</span>, <span class="hljs-number">4963</span>, <span class="hljs-number">1012</span>, <span class="hljs-number">102</span>], | |
| <span class="hljs-string">'token_type_ids'</span>: [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| <span class="hljs-string">'attention_mask'</span>: [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>]}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1gpth0f">Il tokenizer restituisce un dizionario contenente tre oggetti importanti:</p> <ul data-svelte-h="svelte-1j17qe7"><li><a href="glossary#input-ids">input_ids</a> sono gli indici che corrispondono ad ogni token nella frase.</li> <li><a href="glossary#attention-mask">attention_mask</a> indicata se un token deve essere elaborato o no.</li> <li><a href="glossary#token-type-ids">token_type_ids</a> identifica a quale sequenza appartiene un token se è presente più di una sequenza.</li></ul> <p data-svelte-h="svelte-14x83qu">Si possono decodificare gli <code>input_ids</code> per farsi restituire l’input originale:</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-meta">>>> </span>tokenizer.decode(encoded_input[<span class="hljs-string">"input_ids"</span>]) | |
| <span class="hljs-string">'[CLS] Do not meddle in the affairs of wizards, for they are subtle and quick to anger. [SEP]'</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-l2tayd">Come si può vedere, il tokenizer aggiunge due token speciali - <code>CLS</code> e <code>SEP</code> (classificatore e separatore) - alla frase. Non tutti i modelli hanno bisogno dei token speciali, ma se servono, il tokenizer li aggiungerà automaticamente.</p> <p data-svelte-h="svelte-1veh7mu">Se ci sono più frasi che vuoi processare, passale come una lista al tokenizer:</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-meta">>>> </span>batch_sentences = [ | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"But what about second breakfast?"</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"Don't think he knows about second breakfast, Pip."</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"What about elevensies?"</span>, | |
| <span class="hljs-meta">... </span>] | |
| <span class="hljs-meta">>>> </span>encoded_inputs = tokenizer(batch_sentences) | |
| <span class="hljs-meta">>>> </span><span class="hljs-built_in">print</span>(encoded_inputs) | |
| {<span class="hljs-string">'input_ids'</span>: [[<span class="hljs-number">101</span>, <span class="hljs-number">1252</span>, <span class="hljs-number">1184</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">1248</span>, <span class="hljs-number">6462</span>, <span class="hljs-number">136</span>, <span class="hljs-number">102</span>], | |
| [<span class="hljs-number">101</span>, <span class="hljs-number">1790</span>, <span class="hljs-number">112</span>, <span class="hljs-number">189</span>, <span class="hljs-number">1341</span>, <span class="hljs-number">1119</span>, <span class="hljs-number">3520</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">1248</span>, <span class="hljs-number">6462</span>, <span class="hljs-number">117</span>, <span class="hljs-number">21902</span>, <span class="hljs-number">1643</span>, <span class="hljs-number">119</span>, <span class="hljs-number">102</span>], | |
| [<span class="hljs-number">101</span>, <span class="hljs-number">1327</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">5450</span>, <span class="hljs-number">23434</span>, <span class="hljs-number">136</span>, <span class="hljs-number">102</span>]], | |
| <span class="hljs-string">'token_type_ids'</span>: [[<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]], | |
| <span class="hljs-string">'attention_mask'</span>: [[<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>]]}<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="pad" 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="#pad"><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>Pad</span></h3> <p data-svelte-h="svelte-1x5y5yz">Questo è un argomento importante. Quando processi un insieme di frasi potrebbero non avere tutte la stessa lunghezza. Questo è un problema perchè i tensori, in input del modello, devono avere dimensioni uniformi. Il padding è una strategia per assicurarsi che i tensori siano rettangolari aggiungendo uno speciale <em>padding token</em> alle frasi più corte.</p> <p data-svelte-h="svelte-ddtxpr">Imposta il parametro <code>padding</code> a <code>True</code> per imbottire le frasi più corte nel gruppo in modo che combacino con la massima lunghezza presente:</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-meta">>>> </span>batch_sentences = [ | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"But what about second breakfast?"</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"Don't think he knows about second breakfast, Pip."</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"What about elevensies?"</span>, | |
| <span class="hljs-meta">... </span>] | |
| <span class="hljs-meta">>>> </span>encoded_input = tokenizer(batch_sentences, padding=<span class="hljs-literal">True</span>) | |
| <span class="hljs-meta">>>> </span><span class="hljs-built_in">print</span>(encoded_input) | |
| {<span class="hljs-string">'input_ids'</span>: [[<span class="hljs-number">101</span>, <span class="hljs-number">1252</span>, <span class="hljs-number">1184</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">1248</span>, <span class="hljs-number">6462</span>, <span class="hljs-number">136</span>, <span class="hljs-number">102</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">101</span>, <span class="hljs-number">1790</span>, <span class="hljs-number">112</span>, <span class="hljs-number">189</span>, <span class="hljs-number">1341</span>, <span class="hljs-number">1119</span>, <span class="hljs-number">3520</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">1248</span>, <span class="hljs-number">6462</span>, <span class="hljs-number">117</span>, <span class="hljs-number">21902</span>, <span class="hljs-number">1643</span>, <span class="hljs-number">119</span>, <span class="hljs-number">102</span>], | |
| [<span class="hljs-number">101</span>, <span class="hljs-number">1327</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">5450</span>, <span class="hljs-number">23434</span>, <span class="hljs-number">136</span>, <span class="hljs-number">102</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]], | |
| <span class="hljs-string">'token_type_ids'</span>: [[<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]], | |
| <span class="hljs-string">'attention_mask'</span>: [[<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]]}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-prnmcs">Nota che il tokenizer aggiunge alle sequenze degli <code>0</code> perchè sono troppo corte!</p> <h3 class="relative group"><a id="truncation" 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="#truncation"><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>Truncation</span></h3> <p data-svelte-h="svelte-ptwyk1">L’altra faccia della medaglia è che avolte le sequenze possono essere troppo lunghe per essere gestite dal modello. In questo caso, avrai bisogno di troncare la sequenza per avere una lunghezza minore.</p> <p data-svelte-h="svelte-j9uhec">Imposta il parametro <code>truncation</code> a <code>True</code> per troncare una sequenza alla massima lunghezza accettata dal modello:</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-meta">>>> </span>batch_sentences = [ | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"But what about second breakfast?"</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"Don't think he knows about second breakfast, Pip."</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"What about elevensies?"</span>, | |
| <span class="hljs-meta">... </span>] | |
| <span class="hljs-meta">>>> </span>encoded_input = tokenizer(batch_sentences, padding=<span class="hljs-literal">True</span>, truncation=<span class="hljs-literal">True</span>) | |
| <span class="hljs-meta">>>> </span><span class="hljs-built_in">print</span>(encoded_input) | |
| {<span class="hljs-string">'input_ids'</span>: [[<span class="hljs-number">101</span>, <span class="hljs-number">1252</span>, <span class="hljs-number">1184</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">1248</span>, <span class="hljs-number">6462</span>, <span class="hljs-number">136</span>, <span class="hljs-number">102</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">101</span>, <span class="hljs-number">1790</span>, <span class="hljs-number">112</span>, <span class="hljs-number">189</span>, <span class="hljs-number">1341</span>, <span class="hljs-number">1119</span>, <span class="hljs-number">3520</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">1248</span>, <span class="hljs-number">6462</span>, <span class="hljs-number">117</span>, <span class="hljs-number">21902</span>, <span class="hljs-number">1643</span>, <span class="hljs-number">119</span>, <span class="hljs-number">102</span>], | |
| [<span class="hljs-number">101</span>, <span class="hljs-number">1327</span>, <span class="hljs-number">1164</span>, <span class="hljs-number">5450</span>, <span class="hljs-number">23434</span>, <span class="hljs-number">136</span>, <span class="hljs-number">102</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]], | |
| <span class="hljs-string">'token_type_ids'</span>: [[<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]], | |
| <span class="hljs-string">'attention_mask'</span>: [[<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]]}<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="costruire-i-tensori" 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="#costruire-i-tensori"><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>Costruire i tensori</span></h3> <p data-svelte-h="svelte-1x9x7br">Infine, vuoi che il tokenizer restituisca i tensori prodotti dal modello.</p> <p data-svelte-h="svelte-1uhzoe4">Imposta il parametro <code>return_tensors</code> su <code>pt</code> per PyTorch, o <code>tf</code> per TensorFlow:</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="Copy code excerpt to clipboard" 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=""><div><div class="bg-white leading-none border border-gray-100 rounded-lg inline-flex p-0.5 text-sm mb-4 select-none"><button class="flex justify-center py-1.5 px-2.5 focus:outline-none rounded-l false"><svg class="mr-1.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><defs><clipPath id="a"><rect x="3.05" y="0.5" width="25.73" height="31" fill="none"></rect></clipPath></defs><g clip-path="url(#a)"><path d="M24.94,9.51a12.81,12.81,0,0,1,0,18.16,12.68,12.68,0,0,1-18,0,12.81,12.81,0,0,1,0-18.16l9-9V5l-.84.83-6,6a9.58,9.58,0,1,0,13.55,0ZM20.44,9a1.68,1.68,0,1,1,1.67-1.67A1.68,1.68,0,0,1,20.44,9Z" fill="#ee4c2c"></path></g></svg> <p class="!m-0 ">Pytorch</p> </button><button class="flex justify-center py-1.5 px-2.5 focus:outline-none rounded-r text-gray-500 filter grayscale"><svg class="mr-1.5" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" focusable="false" role="img" width="0.94em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 274"><path d="M145.726 42.065v42.07l72.861 42.07v-42.07l-72.86-42.07zM0 84.135v42.07l36.43 21.03V105.17L0 84.135zm109.291 21.035l-36.43 21.034v126.2l36.43 21.035v-84.135l36.435 21.035v-42.07l-36.435-21.034V105.17z" fill="#E55B2D"></path><path d="M145.726 42.065L36.43 105.17v42.065l72.861-42.065v42.065l36.435-21.03v-84.14zM255.022 63.1l-36.435 21.035v42.07l36.435-21.035V63.1zm-72.865 84.135l-36.43 21.035v42.07l36.43-21.036v-42.07zm-36.43 63.104l-36.436-21.035v84.135l36.435-21.035V210.34z" fill="#ED8E24"></path><path d="M145.726 0L0 84.135l36.43 21.035l109.296-63.105l72.861 42.07L255.022 63.1L145.726 0zm0 126.204l-36.435 21.03l36.435 21.036l36.43-21.035l-36.43-21.03z" fill="#F8BF3C"></path></svg> <p class="!m-0 ">TensorFlow</p> </button></div></div><!-- HTML_TAG_START --><span class="hljs-meta">>>> </span>batch_sentences = [ | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"But what about second breakfast?"</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"Don't think he knows about second breakfast, Pip."</span>, | |
| <span class="hljs-meta">... </span> <span class="hljs-string">"What about elevensies?"</span>, | |
| <span class="hljs-meta">... </span>] | |
| <span class="hljs-meta">>>> </span>encoded_input = tokenizer(batch, padding=<span class="hljs-literal">True</span>, truncation=<span class="hljs-literal">True</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| <span class="hljs-meta">>>> </span><span class="hljs-built_in">print</span>(encoded_input) | |
| {<span class="hljs-string">'input_ids'</span>: tensor([[ <span class="hljs-number">101</span>, <span class="hljs-number">153</span>, <span class="hljs-number">7719</span>, <span class="hljs-number">21490</span>, <span class="hljs-number">1122</span>, <span class="hljs-number">1114</span>, <span class="hljs-number">9582</span>, <span class="hljs-number">1623</span>, <span class="hljs-number">102</span>], | |
| [ <span class="hljs-number">101</span>, <span class="hljs-number">5226</span>, <span class="hljs-number">1122</span>, <span class="hljs-number">9649</span>, <span class="hljs-number">1199</span>, <span class="hljs-number">2610</span>, <span class="hljs-number">1236</span>, <span class="hljs-number">102</span>, <span class="hljs-number">0</span>]]), | |
| <span class="hljs-string">'token_type_ids'</span>: tensor([[<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>], | |
| [<span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>, <span class="hljs-number">0</span>]]), | |
| <span class="hljs-string">'attention_mask'</span>: tensor([[<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>], | |
| [<span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>, <span class="hljs-number">0</span>]])}<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="audio" 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="#audio"><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>Audio</span></h2> <p data-svelte-h="svelte-tx36aq">Gli input audio sono processati in modo differente rispetto al testo, ma l’obiettivo rimane lo stesso: creare sequenze numeriche che il modello può capire. Un <a href="main_classes/feature_extractor">estrattore di caratteristiche</a> è progettato con lo scopo preciso di estrarre caratteristiche da immagini o dati audio grezzi e convertirli in tensori. Prima di iniziare, installa 🤗 Datasets per caricare un dataset audio e sperimentare:</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 -->pip install datasets<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1jtnw4g">Carica il dataset <a href="https://huggingface.co/datasets/PolyAI/minds14" rel="nofollow">MInDS-14</a> (vedi il 🤗 <a href="https://huggingface.co/docs/datasets/load_hub" rel="nofollow">Datasets tutorial</a> per avere maggiori dettagli su come caricare un dataset):</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-meta">>>> </span><span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset, Audio | |
| <span class="hljs-meta">>>> </span>dataset = load_dataset(<span class="hljs-string">"PolyAI/minds14"</span>, name=<span class="hljs-string">"en-US"</span>, split=<span class="hljs-string">"train"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-m0eu7r">Accedi al primo elemento della colonna <code>audio</code> per dare uno sguardo all’input. Richiamando la colonna <code>audio</code> sarà caricato automaticamente e ricampionato il file audio:</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-meta">>>> </span>dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"audio"</span>] | |
| {<span class="hljs-string">'array'</span>: array([ <span class="hljs-number">0.</span> , <span class="hljs-number">0.00024414</span>, -<span class="hljs-number">0.00024414</span>, ..., -<span class="hljs-number">0.00024414</span>, | |
| <span class="hljs-number">0.</span> , <span class="hljs-number">0.</span> ], dtype=float32), | |
| <span class="hljs-string">'path'</span>: <span class="hljs-string">'/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav'</span>, | |
| <span class="hljs-string">'sampling_rate'</span>: <span class="hljs-number">8000</span>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-yck6ch">Questo restituisce tre oggetti:</p> <ul data-svelte-h="svelte-1253maj"><li><code>array</code> è il segnale vocale caricato - e potenzialmente ricampionato - come vettore 1D.</li> <li><code>path</code> il percorso del file audio.</li> <li><code>sampling_rate</code> si riferisce al numero di campioni del segnale vocale misurati al secondo.</li></ul> <h3 class="relative group"><a id="ricampionamento" 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="#ricampionamento"><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>Ricampionamento</span></h3> <p data-svelte-h="svelte-11bqj3i">Per questo tutorial, puoi usare il modello <a href="https://huggingface.co/facebook/wav2vec2-base" rel="nofollow">Wav2Vec2</a>. Come puoi vedere dalla model card, il modello Wav2Vec2 è preaddestrato su un campionamento vocale a 16kHz.È importante che la frequenza di campionamento dei tuoi dati audio combaci con la frequenza di campionamento del dataset usato per preaddestrare il modello. Se la frequenza di campionamento dei tuoi dati non è uguale dovrai ricampionare i tuoi dati audio.</p> <p data-svelte-h="svelte-1t3fg9q">Per esempio, il dataset <a href="https://huggingface.co/datasets/PolyAI/minds14" rel="nofollow">MInDS-14</a> ha una frequenza di campionamento di 8000kHz. Utilizzando il modello Wav2Vec2 su questo dataset, alzala a 16kHz:</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-meta">>>> </span>dataset = load_dataset(<span class="hljs-string">"PolyAI/minds14"</span>, name=<span class="hljs-string">"en-US"</span>, split=<span class="hljs-string">"train"</span>) | |
| <span class="hljs-meta">>>> </span>dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"audio"</span>] | |
| {<span class="hljs-string">'array'</span>: array([ <span class="hljs-number">0.</span> , <span class="hljs-number">0.00024414</span>, -<span class="hljs-number">0.00024414</span>, ..., -<span class="hljs-number">0.00024414</span>, | |
| <span class="hljs-number">0.</span> , <span class="hljs-number">0.</span> ], dtype=float32), | |
| <span class="hljs-string">'path'</span>: <span class="hljs-string">'/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav'</span>, | |
| <span class="hljs-string">'sampling_rate'</span>: <span class="hljs-number">8000</span>}<!-- HTML_TAG_END --></pre></div> <ol data-svelte-h="svelte-ofmzcd"><li>Usa il metodo di 🤗 Datasets’ <a href="https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.Dataset.cast_column" rel="nofollow"><code>cast_column</code></a> per alzare la frequenza di campionamento a 16kHz:</li></ol> <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-meta">>>> </span>dataset = dataset.cast_column(<span class="hljs-string">"audio"</span>, Audio(sampling_rate=<span class="hljs-number">16_000</span>))<!-- HTML_TAG_END --></pre></div> <ol start="2" data-svelte-h="svelte-jojg0k"><li>Carica il file audio:</li></ol> <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-meta">>>> </span>dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"audio"</span>] | |
| {<span class="hljs-string">'array'</span>: array([ <span class="hljs-number">2.3443763e-05</span>, <span class="hljs-number">2.1729663e-04</span>, <span class="hljs-number">2.2145823e-04</span>, ..., | |
| <span class="hljs-number">3.8356509e-05</span>, -<span class="hljs-number">7.3497440e-06</span>, -<span class="hljs-number">2.1754686e-05</span>], dtype=float32), | |
| <span class="hljs-string">'path'</span>: <span class="hljs-string">'/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav'</span>, | |
| <span class="hljs-string">'sampling_rate'</span>: <span class="hljs-number">16000</span>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1bt1o9m">Come puoi notare, la <code>sampling_rate</code> adesso è 16kHz!</p> <h3 class="relative group"><a id="feature-extractor" 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="#feature-extractor"><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>Feature extractor</span></h3> <p data-svelte-h="svelte-1e8d4ks">Il prossimo passo è caricare un estrattore di caratteristiche per normalizzare e fare padding sull’input. Quando applichiamo il padding sui dati testuali, uno <code>0</code> è aggiunto alle sequenze più brevi. La stessa idea si applica ai dati audio, l’estrattore di caratteristiche per gli audio aggiungerà uno <code>0</code> - interpretato come silenzio - agli <code>array</code>.</p> <p data-svelte-h="svelte-87e7to">Carica l’estrattore delle caratteristiche con <code>AutoFeatureExtractor.from_pretrained()</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 --><span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoFeatureExtractor | |
| <span class="hljs-meta">>>> </span>feature_extractor = AutoFeatureExtractor.from_pretrained(<span class="hljs-string">"facebook/wav2vec2-base"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-pj9uvo">Inserisci l’ <code>array</code> audio nell’estrattore delle caratteristiche. Noi raccomandiamo sempre di aggiungere il parametro <code>sampling_rate</code> nell’estrattore delle caratteristiche per correggere meglio qualche errore, dovuto ai silenzi, che potrebbe verificarsi.</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-meta">>>> </span>audio_input = [dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"audio"</span>][<span class="hljs-string">"array"</span>]] | |
| <span class="hljs-meta">>>> </span>feature_extractor(audio_input, sampling_rate=<span class="hljs-number">16000</span>) | |
| {<span class="hljs-string">'input_values'</span>: [array([ <span class="hljs-number">3.8106556e-04</span>, <span class="hljs-number">2.7506407e-03</span>, <span class="hljs-number">2.8015103e-03</span>, ..., | |
| <span class="hljs-number">5.6335266e-04</span>, <span class="hljs-number">4.6588284e-06</span>, -<span class="hljs-number">1.7142107e-04</span>], dtype=float32)]}<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="pad-e-truncate" 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="#pad-e-truncate"><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>Pad e truncate</span></h3> <p data-svelte-h="svelte-57wuaq">Come per il tokenizer, puoi applicare le operazioni padding o truncation per manipolare sequenze di variabili a lotti. Dai uno sguaro alla lunghezza delle sequenze di questi due campioni audio:</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-meta">>>> </span>dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"audio"</span>][<span class="hljs-string">"array"</span>].shape | |
| (<span class="hljs-number">173398</span>,) | |
| <span class="hljs-meta">>>> </span>dataset[<span class="hljs-number">1</span>][<span class="hljs-string">"audio"</span>][<span class="hljs-string">"array"</span>].shape | |
| (<span class="hljs-number">106496</span>,)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-4umahi">Come puoi vedere, il primo campione ha una sequenza più lunga del secondo. Crea una funzione che preprocesserà il dataset. Specifica una lunghezza massima del campione, e l’estrattore di features si occuperà di riempire o troncare la sequenza per coincidervi:</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-meta">>>> </span><span class="hljs-keyword">def</span> <span class="hljs-title function_">preprocess_function</span>(<span class="hljs-params">examples</span>): | |
| <span class="hljs-meta">... </span> audio_arrays = [x[<span class="hljs-string">"array"</span>] <span class="hljs-keyword">for</span> x <span class="hljs-keyword">in</span> examples[<span class="hljs-string">"audio"</span>]] | |
| <span class="hljs-meta">... </span> inputs = feature_extractor( | |
| <span class="hljs-meta">... </span> audio_arrays, | |
| <span class="hljs-meta">... </span> sampling_rate=<span class="hljs-number">16000</span>, | |
| <span class="hljs-meta">... </span> padding=<span class="hljs-literal">True</span>, | |
| <span class="hljs-meta">... </span> max_length=<span class="hljs-number">100000</span>, | |
| <span class="hljs-meta">... </span> truncation=<span class="hljs-literal">True</span>, | |
| <span class="hljs-meta">... </span> ) | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">return</span> inputs<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-g1tkoi">Applica la funzione ai primi esempi nel dataset:</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-meta">>>> </span>processed_dataset = preprocess_function(dataset[:<span class="hljs-number">5</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-zko717">Adesso guarda la lunghezza dei campioni elaborati:</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-meta">>>> </span>processed_dataset[<span class="hljs-string">"input_values"</span>][<span class="hljs-number">0</span>].shape | |
| (<span class="hljs-number">100000</span>,) | |
| <span class="hljs-meta">>>> </span>processed_dataset[<span class="hljs-string">"input_values"</span>][<span class="hljs-number">1</span>].shape | |
| (<span class="hljs-number">100000</span>,)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1w2gql8">La lunghezza dei campioni adesso coincide con la massima lunghezza impostata nelle funzione.</p> <h2 class="relative group"><a id="vision" 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="#vision"><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>Vision</span></h2> <p data-svelte-h="svelte-d9w7z3">Un estrattore di caratteristiche si può usare anche per processare immagini e per compiti di visione. Ancora una volta, l’obiettivo è convertire l’immagine grezza in un lotto di tensori come input.</p> <p data-svelte-h="svelte-1ed7kvz">Carica il dataset <a href="https://huggingface.co/datasets/food101" rel="nofollow">food101</a> per questa esercitazione. Usa il parametro <code>split</code> di 🤗 Datasets per caricare solo un piccolo campione dal dataset di addestramento poichè il set di dati è molto grande:</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-meta">>>> </span><span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset | |
| <span class="hljs-meta">>>> </span>dataset = load_dataset(<span class="hljs-string">"food101"</span>, split=<span class="hljs-string">"train[:100]"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-nr0m9j">Secondo passo, dai uno sguardo alle immagini usando la caratteristica <a href="https://huggingface.co/docs/datasets/package_reference/main_classes.html?highlight=image#datasets.Image" rel="nofollow"><code>Image</code></a> di 🤗 Datasets:</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-meta">>>> </span>dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"image"</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1pswjfa"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/vision-preprocess-tutorial.png" alt="vision-preprocess-tutorial.png"></p> <h3 class="relative group"><a id="feature-extractor" 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="#feature-extractor"><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>Feature extractor</span></h3> <p data-svelte-h="svelte-hrvdft">Carica l’estrattore di caratteristiche <code>AutoFeatureExtractor.from_pretrained()</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 --><span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoFeatureExtractor | |
| <span class="hljs-meta">>>> </span>feature_extractor = AutoFeatureExtractor.from_pretrained(<span class="hljs-string">"google/vit-base-patch16-224"</span>)<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="data-augmentation" 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="#data-augmentation"><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>Data augmentation</span></h3> <p data-svelte-h="svelte-1v0wu38">Per le attività di visione, è usuale aggiungere alcuni tipi di data augmentation alle immagini come parte del preprocessing. Puoi aggiungere augmentations con qualsiasi libreria che preferisci, ma in questa esercitazione, userai il modulo <a href="https://pytorch.org/vision/stable/transforms.html" rel="nofollow"><code>transforms</code></a> di torchvision.</p> <ol data-svelte-h="svelte-1echh8a"><li>Normalizza l’immagine e usa <a href="https://pytorch.org/vision/master/generated/torchvision.transforms.Compose.html" rel="nofollow"><code>Compose</code></a> per concatenare alcune trasformazioni - <a href="https://pytorch.org/vision/main/generated/torchvision.transforms.RandomResizedCrop.html" rel="nofollow"><code>RandomResizedCrop</code></a> e <a href="https://pytorch.org/vision/main/generated/torchvision.transforms.ColorJitter.html" rel="nofollow"><code>ColorJitter</code></a> - insieme:</li></ol> <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-meta">>>> </span><span class="hljs-keyword">from</span> torchvision.transforms <span class="hljs-keyword">import</span> Compose, Normalize, RandomResizedCrop, ColorJitter, ToTensor | |
| <span class="hljs-meta">>>> </span>normalize = Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std) | |
| <span class="hljs-meta">>>> </span>_transforms = Compose( | |
| <span class="hljs-meta">... </span> [RandomResizedCrop(feature_extractor.size), ColorJitter(brightness=<span class="hljs-number">0.5</span>, hue=<span class="hljs-number">0.5</span>), ToTensor(), normalize] | |
| <span class="hljs-meta">... </span>)<!-- HTML_TAG_END --></pre></div> <ol start="2" data-svelte-h="svelte-1nbnku8"><li>Il modello accetta <a href="model_doc/visionencoderdecoder#transformers.VisionEncoderDecoderModel.forward.pixel_values"><code>pixel_values</code></a> come input. Questo valore è generato dall’estrattore di caratteristiche. Crea una funzione che genera <code>pixel_values</code> dai transforms:</li></ol> <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-meta">>>> </span><span class="hljs-keyword">def</span> <span class="hljs-title function_">transforms</span>(<span class="hljs-params">examples</span>): | |
| <span class="hljs-meta">... </span> examples[<span class="hljs-string">"pixel_values"</span>] = [_transforms(image.convert(<span class="hljs-string">"RGB"</span>)) <span class="hljs-keyword">for</span> image <span class="hljs-keyword">in</span> examples[<span class="hljs-string">"image"</span>]] | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">return</span> examples<!-- HTML_TAG_END --></pre></div> <ol start="3" data-svelte-h="svelte-13mtvsg"><li>Poi utilizza 🤗 Datasets <a href="https://huggingface.co/docs/datasets/process#format-transform" rel="nofollow"><code>set_transform</code></a>per applicare al volo la trasformazione:</li></ol> <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-meta">>>> </span>dataset.set_transform(transforms)<!-- HTML_TAG_END --></pre></div> <ol start="4" data-svelte-h="svelte-z3uetu"><li>Adesso quando accedi all’immagine, puoi notare che l’estrattore di caratteristiche ha aggiunto <code>pixel_values</code> allo schema di input:</li></ol> <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-meta">>>> </span>dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"image"</span>] | |
| {<span class="hljs-string">'image'</span>: <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=384x512 at <span class="hljs-number">0x7F1A7B0630D0</span>>, | |
| <span class="hljs-string">'label'</span>: <span class="hljs-number">6</span>, | |
| <span class="hljs-string">'pixel_values'</span>: tensor([[[ <span class="hljs-number">0.0353</span>, <span class="hljs-number">0.0745</span>, <span class="hljs-number">0.1216</span>, ..., -<span class="hljs-number">0.9922</span>, -<span class="hljs-number">0.9922</span>, -<span class="hljs-number">0.9922</span>], | |
| [-<span class="hljs-number">0.0196</span>, <span class="hljs-number">0.0667</span>, <span class="hljs-number">0.1294</span>, ..., -<span class="hljs-number">0.9765</span>, -<span class="hljs-number">0.9843</span>, -<span class="hljs-number">0.9922</span>], | |
| [ <span class="hljs-number">0.0196</span>, <span class="hljs-number">0.0824</span>, <span class="hljs-number">0.1137</span>, ..., -<span class="hljs-number">0.9765</span>, -<span class="hljs-number">0.9686</span>, -<span class="hljs-number">0.8667</span>], | |
| ..., | |
| [ <span class="hljs-number">0.0275</span>, <span class="hljs-number">0.0745</span>, <span class="hljs-number">0.0510</span>, ..., -<span class="hljs-number">0.1137</span>, -<span class="hljs-number">0.1216</span>, -<span class="hljs-number">0.0824</span>], | |
| [ <span class="hljs-number">0.0667</span>, <span class="hljs-number">0.0824</span>, <span class="hljs-number">0.0667</span>, ..., -<span class="hljs-number">0.0588</span>, -<span class="hljs-number">0.0745</span>, -<span class="hljs-number">0.0980</span>], | |
| [ <span class="hljs-number">0.0353</span>, <span class="hljs-number">0.0353</span>, <span class="hljs-number">0.0431</span>, ..., -<span class="hljs-number">0.0039</span>, -<span class="hljs-number">0.0039</span>, -<span class="hljs-number">0.0588</span>]], | |
| [[ <span class="hljs-number">0.2078</span>, <span class="hljs-number">0.2471</span>, <span class="hljs-number">0.2863</span>, ..., -<span class="hljs-number">0.9451</span>, -<span class="hljs-number">0.9373</span>, -<span class="hljs-number">0.9451</span>], | |
| [ <span class="hljs-number">0.1608</span>, <span class="hljs-number">0.2471</span>, <span class="hljs-number">0.3098</span>, ..., -<span class="hljs-number">0.9373</span>, -<span class="hljs-number">0.9451</span>, -<span class="hljs-number">0.9373</span>], | |
| [ <span class="hljs-number">0.2078</span>, <span class="hljs-number">0.2706</span>, <span class="hljs-number">0.3020</span>, ..., -<span class="hljs-number">0.9608</span>, -<span class="hljs-number">0.9373</span>, -<span class="hljs-number">0.8275</span>], | |
| ..., | |
| [-<span class="hljs-number">0.0353</span>, <span class="hljs-number">0.0118</span>, -<span class="hljs-number">0.0039</span>, ..., -<span class="hljs-number">0.2392</span>, -<span class="hljs-number">0.2471</span>, -<span class="hljs-number">0.2078</span>], | |
| [ <span class="hljs-number">0.0196</span>, <span class="hljs-number">0.0353</span>, <span class="hljs-number">0.0196</span>, ..., -<span class="hljs-number">0.1843</span>, -<span class="hljs-number">0.2000</span>, -<span class="hljs-number">0.2235</span>], | |
| [-<span class="hljs-number">0.0118</span>, -<span class="hljs-number">0.0039</span>, -<span class="hljs-number">0.0039</span>, ..., -<span class="hljs-number">0.0980</span>, -<span class="hljs-number">0.0980</span>, -<span class="hljs-number">0.1529</span>]], | |
| [[ <span class="hljs-number">0.3961</span>, <span class="hljs-number">0.4431</span>, <span class="hljs-number">0.4980</span>, ..., -<span class="hljs-number">0.9216</span>, -<span class="hljs-number">0.9137</span>, -<span class="hljs-number">0.9216</span>], | |
| [ <span class="hljs-number">0.3569</span>, <span class="hljs-number">0.4510</span>, <span class="hljs-number">0.5216</span>, ..., -<span class="hljs-number">0.9059</span>, -<span class="hljs-number">0.9137</span>, -<span class="hljs-number">0.9137</span>], | |
| [ <span class="hljs-number">0.4118</span>, <span class="hljs-number">0.4745</span>, <span class="hljs-number">0.5216</span>, ..., -<span class="hljs-number">0.9137</span>, -<span class="hljs-number">0.8902</span>, -<span class="hljs-number">0.7804</span>], | |
| ..., | |
| [-<span class="hljs-number">0.2314</span>, -<span class="hljs-number">0.1922</span>, -<span class="hljs-number">0.2078</span>, ..., -<span class="hljs-number">0.4196</span>, -<span class="hljs-number">0.4275</span>, -<span class="hljs-number">0.3882</span>], | |
| [-<span class="hljs-number">0.1843</span>, -<span class="hljs-number">0.1686</span>, -<span class="hljs-number">0.2000</span>, ..., -<span class="hljs-number">0.3647</span>, -<span class="hljs-number">0.3804</span>, -<span class="hljs-number">0.4039</span>], | |
| [-<span class="hljs-number">0.1922</span>, -<span class="hljs-number">0.1922</span>, -<span class="hljs-number">0.1922</span>, ..., -<span class="hljs-number">0.2941</span>, -<span class="hljs-number">0.2863</span>, -<span class="hljs-number">0.3412</span>]]])}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1dqmbof">Di seguito come si vede l’immagine dopo la fase di preprocessing. Come ci si aspetterebbe dalle trasformazioni applicate, l’immagine è stata ritagliata in modo casuale e le proprietà del colore sono diverse.</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-meta">>>> </span><span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np | |
| <span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> matplotlib.pyplot <span class="hljs-keyword">as</span> plt | |
| <span class="hljs-meta">>>> </span>img = dataset[<span class="hljs-number">0</span>][<span class="hljs-string">"pixel_values"</span>] | |
| <span class="hljs-meta">>>> </span>plt.imshow(img.permute(<span class="hljs-number">1</span>, <span class="hljs-number">2</span>, <span class="hljs-number">0</span>))<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-vgp975"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/preprocessed_image.png" alt="preprocessed_image"></p> <h2 class="relative group"><a id="multimodal" 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="#multimodal"><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>Multimodal</span></h2> <p data-svelte-h="svelte-5kk6r7">Per attività multimodali userai una combinazione di tutto quello che hai imparato poco fa e applicherai le tue competenze alla comprensione automatica del parlato (Automatic Speech Recognition - ASR). Questo significa che avrai bisogno di:</p> <ul data-svelte-h="svelte-vwiihi"><li>Un estrattore delle caratteristiche per processare i dati audio.</li> <li>Il Tokenizer per processare i testi.</li></ul> <p data-svelte-h="svelte-1c58wda">Ritorna sul datasere <a href="https://huggingface.co/datasets/lj_speech" rel="nofollow">LJ Speech</a>:</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-meta">>>> </span><span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset | |
| <span class="hljs-meta">>>> </span>lj_speech = load_dataset(<span class="hljs-string">"lj_speech"</span>, split=<span class="hljs-string">"train"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1a7lzey">Visto che sei interessato solo alle colonne <code>audio</code> e <code>text</code>, elimina tutte le altre:</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-meta">>>> </span>lj_speech = lj_speech.<span class="hljs-built_in">map</span>(remove_columns=[<span class="hljs-string">"file"</span>, <span class="hljs-string">"id"</span>, <span class="hljs-string">"normalized_text"</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1ityppm">Adesso guarda le colonne <code>audio</code> e <code>text</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 --><span class="hljs-meta">>>> </span>lj_speech[<span class="hljs-number">0</span>][<span class="hljs-string">"audio"</span>] | |
| {<span class="hljs-string">'array'</span>: array([-<span class="hljs-number">7.3242188e-04</span>, -<span class="hljs-number">7.6293945e-04</span>, -<span class="hljs-number">6.4086914e-04</span>, ..., | |
| <span class="hljs-number">7.3242188e-04</span>, <span class="hljs-number">2.1362305e-04</span>, <span class="hljs-number">6.1035156e-05</span>], dtype=float32), | |
| <span class="hljs-string">'path'</span>: <span class="hljs-string">'/root/.cache/huggingface/datasets/downloads/extracted/917ece08c95cf0c4115e45294e3cd0dee724a1165b7fc11798369308a465bd26/LJSpeech-1.1/wavs/LJ001-0001.wav'</span>, | |
| <span class="hljs-string">'sampling_rate'</span>: <span class="hljs-number">22050</span>} | |
| <span class="hljs-meta">>>> </span>lj_speech[<span class="hljs-number">0</span>][<span class="hljs-string">"text"</span>] | |
| <span class="hljs-string">'Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition'</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-13fdrdw">Ricorda dalla sezione precedente sull’elaborazione dei dati audio, tu dovresti sempre <a href="preprocessing#audio">ricampionare</a> la frequenza di campionamento dei tuoi dati audio per farla coincidere con quella del dataset usato dal modello preaddestrato:</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-meta">>>> </span>lj_speech = lj_speech.cast_column(<span class="hljs-string">"audio"</span>, Audio(sampling_rate=<span class="hljs-number">16_000</span>))<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="processor" 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="#processor"><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>Processor</span></h3> <p data-svelte-h="svelte-15ryd1v">Un processor combina un estrattore di caratteristiche e un tokenizer. Carica un processor con <code>AutoProcessor.from_pretrained()</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 --><span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoProcessor | |
| <span class="hljs-meta">>>> </span>processor = AutoProcessor.from_pretrained(<span class="hljs-string">"facebook/wav2vec2-base-960h"</span>)<!-- HTML_TAG_END --></pre></div> <ol data-svelte-h="svelte-1axefc1"><li>Crea una funzione che processi i dati audio in <code>input_values</code>, e tokenizza il testo in <code>labels</code>. Questi sono i tuoi input per il modello:</li></ol> <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-meta">>>> </span><span class="hljs-keyword">def</span> <span class="hljs-title function_">prepare_dataset</span>(<span class="hljs-params">example</span>): | |
| <span class="hljs-meta">... </span> audio = example[<span class="hljs-string">"audio"</span>] | |
| <span class="hljs-meta">... </span> example.update(processor(audio=audio[<span class="hljs-string">"array"</span>], text=example[<span class="hljs-string">"text"</span>], sampling_rate=<span class="hljs-number">16000</span>)) | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">return</span> example<!-- HTML_TAG_END --></pre></div> <ol start="2" data-svelte-h="svelte-uxnp00"><li>Applica la funzione <code>prepare_dataset</code> ad un campione:</li></ol> <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-meta">>>> </span>prepare_dataset(lj_speech[<span class="hljs-number">0</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-cbpe6h">Nota che il processor ha aggiunto <code>input_values</code> e <code>labels</code>. La frequenza di campionamento è stata corretta riducendola a 16kHz.</p> <p data-svelte-h="svelte-10fl42v">Fantastico, ora dovresti essere in grado di preelaborare i dati per qualsiasi modalità e persino di combinare modalità diverse! Nella prossima esercitazione, impareremo a mettere a punto un modello sui dati appena pre-elaborati.</p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/transformers/blob/main/docs/source/it/preprocessing.md" 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_728xau = { | |
| assets: "/docs/transformers/pr_33913/it", | |
| base: "/docs/transformers/pr_33913/it", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/transformers/pr_33913/it/_app/immutable/entry/start.cef8f86b.js"), | |
| import("/docs/transformers/pr_33913/it/_app/immutable/entry/app.d399d34f.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 28], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 122 kB
- Xet hash:
- 19c96c76fa01f6751e70918f903521be2151957018578eb4c313ee0db43cc2c9
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.