Buckets:

hf-doc-build/doc / transformers /v4.22.2 /pt /fast_tokenizers.html
rtrm's picture
download
raw
13.9 kB
<meta charset="utf-8" /><meta http-equiv="content-security-policy" content=""><meta name="hf:doc:metadata" content="{&quot;local&quot;:&quot;usando-os-tokenizers-do-tokenizers&quot;,&quot;sections&quot;:[{&quot;local&quot;:&quot;carregando-diretamente-de-um-objeto-tokenizer&quot;,&quot;title&quot;:&quot;Carregando diretamente de um objeto tokenizer&quot;},{&quot;local&quot;:&quot;carregando-de-um-arquivo-json&quot;,&quot;title&quot;:&quot;Carregando de um arquivo JSON&quot;}],&quot;title&quot;:&quot;Usando os Tokenizers do 🤗 Tokenizers&quot;}" data-svelte="svelte-1phssyn">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/assets/pages/__layout.svelte-hf-doc-builder.css">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/start-hf-doc-builder.js">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/chunks/vendor-hf-doc-builder.js">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/chunks/paths-hf-doc-builder.js">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/pages/__layout.svelte-hf-doc-builder.js">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/pages/fast_tokenizers.mdx-hf-doc-builder.js">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/chunks/IconCopyLink-hf-doc-builder.js">
<link rel="modulepreload" href="/docs/transformers/v4.22.2/pt/_app/chunks/CodeBlock-hf-doc-builder.js">
<h1 class="relative group"><a id="usando-os-tokenizers-do-tokenizers" 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="#usando-os-tokenizers-do-tokenizers"><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>Usando os Tokenizers do 🤗 Tokenizers
</span></h1>
<p>O <code>PreTrainedTokenizerFast</code> depende da biblioteca <a href="https://huggingface.co/docs/tokenizers" rel="nofollow">🤗 Tokenizers</a>. O Tokenizer obtido da biblioteca 🤗 Tokenizers pode ser carregado facilmente pelo 🤗 Transformers.</p>
<p>Antes de entrar nos detalhes, vamos começar criando um tokenizer fictício em algumas linhas:</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><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> tokenizers <span class="hljs-keyword">import</span> Tokenizer
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> tokenizers.models <span class="hljs-keyword">import</span> BPE
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> tokenizers.trainers <span class="hljs-keyword">import</span> BpeTrainer
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> tokenizers.pre_tokenizers <span class="hljs-keyword">import</span> Whitespace
<span class="hljs-meta">&gt;&gt;&gt; </span>tokenizer = Tokenizer(BPE(unk_token=<span class="hljs-string">&quot;[UNK]&quot;</span>))
<span class="hljs-meta">&gt;&gt;&gt; </span>trainer = BpeTrainer(special_tokens=[<span class="hljs-string">&quot;[UNK]&quot;</span>, <span class="hljs-string">&quot;[CLS]&quot;</span>, <span class="hljs-string">&quot;[SEP]&quot;</span>, <span class="hljs-string">&quot;[PAD]&quot;</span>, <span class="hljs-string">&quot;[MASK]&quot;</span>])
<span class="hljs-meta">&gt;&gt;&gt; </span>tokenizer.pre_tokenizer = Whitespace()
<span class="hljs-meta">&gt;&gt;&gt; </span>files = [...]
<span class="hljs-meta">&gt;&gt;&gt; </span>tokenizer.train(files, trainer)<!-- HTML_TAG_END --></pre></div>
<p>Agora temos um tokenizer treinado nos arquivos que foram definidos. Nós podemos continuar usando nessa execução ou salvar em um arquivo JSON para re-utilizar no futuro.</p>
<h2 class="relative group"><a id="carregando-diretamente-de-um-objeto-tokenizer" 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="#carregando-diretamente-de-um-objeto-tokenizer"><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>Carregando diretamente de um objeto tokenizer
</span></h2>
<p>Vamos ver como aproveitar esse objeto tokenizer na biblioteca 🤗 Transformers. A classe <code>PreTrainedTokenizerFast</code> permite uma instanciação fácil, aceitando o objeto <em>tokenizer</em> instanciado como um argumento:</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><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> PreTrainedTokenizerFast
<span class="hljs-meta">&gt;&gt;&gt; </span>fast_tokenizer = PreTrainedTokenizerFast(tokenizer_object=tokenizer)<!-- HTML_TAG_END --></pre></div>
<p>Esse objeto pode ser utilizado com todos os métodos compartilhados pelos tokenizers dos 🤗 Transformers! Vá para <a href="main_classes/tokenizer">a página do tokenizer</a> para mais informações.</p>
<h2 class="relative group"><a id="carregando-de-um-arquivo-json" 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="#carregando-de-um-arquivo-json"><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>Carregando de um arquivo JSON
</span></h2>
<p>Para carregar um tokenizer de um arquivo JSON vamos primeiro começar salvando nosso 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><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span>tokenizer.save(<span class="hljs-string">&quot;tokenizer.json&quot;</span>)<!-- HTML_TAG_END --></pre></div>
<p>A pasta para qual salvamos esse arquivo pode ser passada para o método de inicialização do <code>PreTrainedTokenizerFast</code> usando o <code>tokenizer_file</code> parâmetro:</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><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> PreTrainedTokenizerFast
<span class="hljs-meta">&gt;&gt;&gt; </span>fast_tokenizer = PreTrainedTokenizerFast(tokenizer_file=<span class="hljs-string">&quot;tokenizer.json&quot;</span>)<!-- HTML_TAG_END --></pre></div>
<p>Esse objeto pode ser utilizado com todos os métodos compartilhados pelos tokenizers dos 🤗 Transformers! Vá para <a href="main_classes/tokenizer">a página do tokenizer</a> para mais informações.</p>
<script type="module" data-hydrate="epv1kh">
import { start } from "/docs/transformers/v4.22.2/pt/_app/start-hf-doc-builder.js";
start({
target: document.querySelector('[data-hydrate="epv1kh"]').parentNode,
paths: {"base":"/docs/transformers/v4.22.2/pt","assets":"/docs/transformers/v4.22.2/pt"},
session: {},
route: false,
spa: false,
trailing_slash: "never",
hydrate: {
status: 200,
error: null,
nodes: [
import("/docs/transformers/v4.22.2/pt/_app/pages/__layout.svelte-hf-doc-builder.js"),
import("/docs/transformers/v4.22.2/pt/_app/pages/fast_tokenizers.mdx-hf-doc-builder.js")
],
params: {}
}
});
</script>

Xet Storage Details

Size:
13.9 kB
·
Xet hash:
ed0461a5bb5d950c94b69c5d4a835dc7bfb8f8bc4a0211907b6871e905ba1820

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