Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Tokenizers","local":"tokenizers","sections":[{"title":"Tokenizer classes","local":"tokenizer-classes","sections":[],"depth":2},{"title":"Multimodal tokenizers","local":"multimodal-tokenizers","sections":[],"depth":2},{"title":"Fast tokenizers","local":"fast-tokenizers","sections":[],"depth":2},{"title":"tiktoken","local":"tiktoken","sections":[{"title":"Create a tiktoken tokenizer","local":"create-a-tiktoken-tokenizer","sections":[],"depth":3}],"depth":2},{"title":"Preprocess","local":"preprocess","sections":[{"title":"Special tokens","local":"special-tokens","sections":[],"depth":3},{"title":"Batch tokenization","local":"batch-tokenization","sections":[],"depth":3},{"title":"Padding","local":"padding","sections":[],"depth":3},{"title":"Truncation","local":"truncation","sections":[],"depth":3}],"depth":2}],"depth":1}"> | |
| <link href="/docs/transformers/pr_33892/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/entry/start.b2c4257a.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/scheduler.31fdf58d.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/singletons.9860629f.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/index.252883d5.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/paths.e85c0ec8.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/entry/app.05ef1f97.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/preload-helper.40847a0e.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/index.2f76fdf0.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/nodes/0.ca4aafa4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/nodes/23.fadb1349.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/CopyLLMTxtMenu.ff482081.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.71f274cc.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/IconCopy.ac192424.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/Youtube.e3933a11.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/CodeBlock.ab12f8e1.js"> | |
| <link rel="modulepreload" href="/docs/transformers/pr_33892/en/_app/immutable/chunks/HfOption.fb051768.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Tokenizers","local":"tokenizers","sections":[{"title":"Tokenizer classes","local":"tokenizer-classes","sections":[],"depth":2},{"title":"Multimodal tokenizers","local":"multimodal-tokenizers","sections":[],"depth":2},{"title":"Fast tokenizers","local":"fast-tokenizers","sections":[],"depth":2},{"title":"tiktoken","local":"tiktoken","sections":[{"title":"Create a tiktoken tokenizer","local":"create-a-tiktoken-tokenizer","sections":[],"depth":3}],"depth":2},{"title":"Preprocess","local":"preprocess","sections":[{"title":"Special tokens","local":"special-tokens","sections":[],"depth":3},{"title":"Batch tokenization","local":"batch-tokenization","sections":[],"depth":3},{"title":"Padding","local":"padding","sections":[],"depth":3},{"title":"Truncation","local":"truncation","sections":[],"depth":3}],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 max-sm:gap-0.5 h-6 max-sm:h-5 px-2 max-sm:px-1.5 text-[11px] max-sm:text-[9px] font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0"><svg class="w-3 h-3 max-sm:w-2.5 max-sm:h-2.5" 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></span> <span>Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-6 max-sm:h-5 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible w-3 h-3 max-sm:w-2.5 max-sm:h-2.5 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg></button></div> </div> <h1 class="relative group"><a id="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="#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>Tokenizers</span></h1> <p data-svelte-h="svelte-11r544d">Tokenizers convert text into an array of numbers known as tensors, the inputs to a text model. There are several tokenizer algorithms, but they all share the same purpose. Split text into smaller words or subwords (tokens) according to some rules, and convert them into numbers (input ids). A Transformers tokenizer also returns an attention mask to indicate which tokens should be attended to.</p> <blockquote class="tip" data-svelte-h="svelte-1ynbhg2"><p>Learn about the most popular tokenization algorithms on the <a href="./tokenizer_summary">Summary of the tokenizers</a> doc.</p></blockquote> <p data-svelte-h="svelte-12f02v">Call <a href="/docs/transformers/pr_33892/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.from_pretrained">from_pretrained()</a> to load a tokenizer and its configuration from the Hugging Face <a href="https://hf.co" rel="nofollow">Hub</a> or a local directory. The pretrained tokenizer is saved in a <a href="https://huggingface.co/google/gemma-2-2b/blob/main/tokenizer.model" rel="nofollow">tokenizer.model</a> file with all its associated vocabulary files.</p> <p data-svelte-h="svelte-1v4y8i9">Pass a string of text to the tokenizer to return the input ids and attention mask, and set the framework tensor type to return with the <code>return_tensors</code> parameter.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"google/gemma-2-2b"</span>) | |
| tokenizer(<span class="hljs-string">"We are very happy to show you the 🤗 Transformers library"</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| {<span class="hljs-string">'input_ids'</span>: tensor([[ <span class="hljs-number">2</span>, <span class="hljs-number">1734</span>, <span class="hljs-number">708</span>, <span class="hljs-number">1508</span>, <span class="hljs-number">4915</span>, <span class="hljs-number">577</span>, <span class="hljs-number">1500</span>, <span class="hljs-number">692</span>, <span class="hljs-number">573</span>, | |
| <span class="hljs-number">156808</span>, <span class="hljs-number">128149</span>, <span class="hljs-number">9581</span>, <span class="hljs-number">235265</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>]]) | |
| }<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-fmq7bu">Whichever tokenizer you use, make sure the tokenizer vocabulary is the same as the pretrained models tokenizer vocabulary. This is especially important if you’re using a custom tokenizer with a different vocabulary from the pretrained models tokenizer.</p> <p data-svelte-h="svelte-74r2x9">This guide provides a brief overview of the tokenizer classes and how to preprocess text with it.</p> <h2 class="relative group"><a id="tokenizer-classes" 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="#tokenizer-classes"><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>Tokenizer classes</span></h2> <p data-svelte-h="svelte-2aprx3">All tokenizers inherit from a <a href="/docs/transformers/pr_33892/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase">PreTrainedTokenizerBase</a> class that provides common methods for all tokenizers like <a href="/docs/transformers/pr_33892/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.from_pretrained">from_pretrained()</a> and <a href="/docs/transformers/pr_33892/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.batch_decode">batch_decode()</a>. There are two main tokenizer classes that build on top of the base class.</p> <ul data-svelte-h="svelte-h2ptke"><li><a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizer">PreTrainedTokenizer</a> is a Python implementation, for example <a href="/docs/transformers/pr_33892/en/model_doc/llama2#transformers.LlamaTokenizer">LlamaTokenizer</a>.</li> <li><a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast">PreTrainedTokenizerFast</a> is a fast Rust-based implementation from the <a href="https://hf.co/docs/tokenizers/index" rel="nofollow">Tokenizers</a> library, for example <a href="/docs/transformers/pr_33892/en/model_doc/llama2#transformers.LlamaTokenizerFast">LlamaTokenizerFast</a>.</li></ul> <p data-svelte-h="svelte-1r47n9z">There are two ways you can load a tokenizer, with <a href="/docs/transformers/pr_33892/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a> or a model-specific tokenizer.</p> <div class="flex space-x-2 items-center my-1.5 mr-8 h-7 !pl-0 -mx-3 md:mx-0"><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd border-gray-800 bg-black dark:bg-gray-700 text-white">AutoTokenizer </div><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd text-gray-500 cursor-pointer opacity-90 hover:text-gray-700 dark:hover:text-gray-200 hover:shadow-sm">model-specific tokenizer </div></div> <div class="language-select"><p data-svelte-h="svelte-1qg1km9">The <a href="./model_doc/auto">AutoClass</a> API is a fast and easy way to load a tokenizer without needing to know whether a Python or Rust-based implementation is available. By default, <a href="/docs/transformers/pr_33892/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a> tries to load a fast tokenizer if it’s available, otherwise, it loads the Python implementation.</p> <p data-svelte-h="svelte-17yeukf">Use <a href="/docs/transformers/pr_33892/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.from_pretrained">from_pretrained()</a> to load a 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-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"google/gemma-2-2b"</span>) | |
| tokenizer(<span class="hljs-string">"We are very happy to show you the 🤗 Transformers library."</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| {<span class="hljs-string">'input_ids'</span>: tensor([[ <span class="hljs-number">2</span>, <span class="hljs-number">1734</span>, <span class="hljs-number">708</span>, <span class="hljs-number">1508</span>, <span class="hljs-number">4915</span>, <span class="hljs-number">577</span>, <span class="hljs-number">1500</span>, <span class="hljs-number">692</span>, <span class="hljs-number">573</span>, | |
| <span class="hljs-number">156808</span>, <span class="hljs-number">128149</span>, <span class="hljs-number">9581</span>, <span class="hljs-number">235265</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>]]) | |
| }<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-24gt5v">Load your own tokenizer by passing its vocabulary file to <a href="/docs/transformers/pr_33892/en/model_doc/auto#transformers.AutoTokenizer.from_pretrained">from_pretrained()</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-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"./model_directory/my_vocab_file.txt"</span>)<!-- HTML_TAG_END --></pre></div> </div> <h2 class="relative group"><a id="multimodal-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="#multimodal-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>Multimodal tokenizers</span></h2> <p data-svelte-h="svelte-1wtso2">In addition to text tokens, multimodal tokenizers also holds tokens from other modalities as a part of its attributes for easy access.</p> <p data-svelte-h="svelte-1oh1svk">To add these special tokens to a tokenizer, pass them as a dictionary to the <code>extra_special_tokens</code> parameter in <a href="/docs/transformers/pr_33892/en/model_doc/auto#transformers.AutoTokenizer.from_pretrained">from_pretrained()</a>. The example below adds the <code>image_token</code> to a vision-language model.</p> <p data-svelte-h="svelte-1xu9dqq">Save the tokenizer so you can reuse it with direct access to the <code>image_token</code>, <code>boi_token</code>, and <code>eoi_token</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 -->vision_tokenizer = AutoTokenizer.from_pretrained( | |
| <span class="hljs-string">"llava-hf/llava-1.5-7b-hf"</span>, | |
| extra_special_tokens={<span class="hljs-string">"image_token"</span>: <span class="hljs-string">"<image>"</span>, <span class="hljs-string">"boi_token"</span>: <span class="hljs-string">"<image_start>"</span>, <span class="hljs-string">"eoi_token"</span>: <span class="hljs-string">"<image_end>"</span>} | |
| ) | |
| <span class="hljs-built_in">print</span>(vision_tokenizer.image_token, vision_tokenizer.image_token_id) | |
| (<span class="hljs-string">"<image>"</span>, <span class="hljs-number">32000</span>) | |
| vision_tokenizer.save_pretrained(<span class="hljs-string">"./path/to/tokenizer"</span>)<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="fast-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="#fast-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>Fast tokenizers</span></h2> <iframe class="w-full xl:w-4/6 h-80" src="https://www.youtube-nocookie.com/embed/3umI3tm27Vw" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe> <p data-svelte-h="svelte-bo6b4d"><a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast">PreTrainedTokenizerFast</a> or <em>fast tokenizers</em> are Rust-based tokenizers from the <a href="https://hf.co/docs/tokenizers" rel="nofollow">Tokenizers</a> library. It is significantly faster at batched tokenization and provides additional alignment methods compared to the Python-based tokenizers.</p> <p data-svelte-h="svelte-qp41iw"><a href="/docs/transformers/pr_33892/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a> automatically loads a fast tokenizer if it’s supported. Otherwise, you need to explicitly load the fast tokenizer.</p> <p data-svelte-h="svelte-1v32tzh">This section will show you how to train a fast tokenizer and reuse it in Transformers.</p> <p data-svelte-h="svelte-1vl9v8s">To train a Byte-Pair Encoding (BPE) tokenizer, create a <a href="https://huggingface.co/docs/tokenizers/main/en/api/tokenizer#tokenizers.Tokenizer" rel="nofollow">Tokenizer</a> and <a href="https://huggingface.co/docs/tokenizers/main/en/api/trainers#tokenizers.trainers.BpeTrainer" rel="nofollow">BpeTrainer</a> class and define the unknown token and special tokens.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> tokenizers <span class="hljs-keyword">import</span> Tokenizer | |
| <span class="hljs-keyword">from</span> tokenizers.models <span class="hljs-keyword">import</span> BPE | |
| <span class="hljs-keyword">from</span> tokenizers.trainers <span class="hljs-keyword">import</span> BpeTrainer | |
| tokenizer = Tokenizer(BPE(unk_token=<span class="hljs-string">"[UNK]"</span>)) | |
| trainer = BpeTrainer(special_tokens=[<span class="hljs-string">"[UNK]"</span>, <span class="hljs-string">"[CLS]"</span>, <span class="hljs-string">"[SEP]"</span>, <span class="hljs-string">"[PAD]"</span>, <span class="hljs-string">"[MASK]"</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-34i0gi">Split the tokens on <a href="https://huggingface.co/docs/tokenizers/main/en/api/pre-tokenizers#tokenizers.pre_tokenizers.Whitespace" rel="nofollow">Whitespace</a> to create tokens that don’t overlap with each other.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> tokenizers.pre_tokenizers <span class="hljs-keyword">import</span> Whitespace | |
| tokenizer.pre_tokenizer = Whitespace()<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ylch61">Call <a href="https://huggingface.co/docs/tokenizers/main/en/api/tokenizer#tokenizers.Tokenizer.train" rel="nofollow">train</a> on the text files and trainer to start training.</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 -->files = [...] | |
| tokenizer.train(files, trainer)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-geyal3">Use <a href="https://huggingface.co/docs/tokenizers/main/en/api/tokenizer#tokenizers.Tokenizer.save" rel="nofollow">save</a> to save the tokenizers configuration and vocabulary to a JSON file.</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 -->tokenizer.save(<span class="hljs-string">"tokenizer.json"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-15n380a">Now you can load and reuse the tokenizer object in Transformers by passing it to the <code>tokenizer_object</code> parameter in <a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast">PreTrainedTokenizerFast</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-keyword">from</span> transformers <span class="hljs-keyword">import</span> PreTrainedTokenizerFast | |
| fast_tokenizer = PreTrainedTokenizerFast(tokenizer_object=tokenizer)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-12ku7k7">To load a saved tokenizer from its JSON file, pass the file path to the <code>tokenizer_file</code> parameter in <a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast">PreTrainedTokenizerFast</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-keyword">from</span> transformers <span class="hljs-keyword">import</span> PreTrainedTokenizerFast | |
| fast_tokenizer = PreTrainedTokenizerFast(tokenizer_file=<span class="hljs-string">"tokenizer.json"</span>)<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="tiktoken" 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="#tiktoken"><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>tiktoken</span></h2> <p data-svelte-h="svelte-10w7qr0"><a href="https://github.com/openai/tiktoken" rel="nofollow">tiktoken</a> is a <a href="./tokenizer_summary#byte-pair-encoding-bpe">byte-pair encoding (BPE)</a> tokenizer by OpenAI. It includes several tokenization schemes or encodings for how text should be tokenized.</p> <p data-svelte-h="svelte-z3kr3y">There are currently two models trained and released with tiktoken, GPT2 and Llama3. Transformers supports models with a <a href="https://hf.co/meta-llama/Meta-Llama-3-8B/blob/main/original/tokenizer.model" rel="nofollow">tokenizer.model</a> tiktoken file. The tiktoken file is automatically converted into Transformers Rust-based <a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast">PreTrainedTokenizerFast</a>.</p> <p data-svelte-h="svelte-9sin8q">Add the <code>subfolder</code> parameter to <a href="/docs/transformers/pr_33892/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> to specify where the <code>tokenizer.model</code> tiktoken file is located.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"meta-llama/Meta-Llama-3-8B-Instruct"</span>, subfolder=<span class="hljs-string">"original"</span>)<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="create-a-tiktoken-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="#create-a-tiktoken-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>Create a tiktoken tokenizer</span></h3> <p data-svelte-h="svelte-fcyep0">The tiktoken <code>tokenizer.model</code> file contains no information about additional tokens or pattern strings. If these are important, convert the tokenizer to <code>tokenizer.json</code> (the appropriate format for <a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast">PreTrainedTokenizerFast</a>).</p> <p data-svelte-h="svelte-t11i5b">Generate the tiktoken <code>tokenizer.model</code> file with the <a href="https://github.com/openai/tiktoken/blob/63527649963def8c759b0f91f2eb69a40934e468/tiktoken/registry.py#L63" rel="nofollow">tiktoken.get_encoding</a> function, and convert it to <code>tokenizer.json</code> with <a href="https://github.com/huggingface/transformers/blob/99e0ab6ed888136ea4877c6d8ab03690a1478363/src/transformers/integrations/tiktoken.py#L8" rel="nofollow">convert_tiktoken_to_fast</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-keyword">from</span> transformers.integrations.tiktoken <span class="hljs-keyword">import</span> convert_tiktoken_to_fast | |
| <span class="hljs-keyword">from</span> tiktoken <span class="hljs-keyword">import</span> get_encoding | |
| <span class="hljs-comment"># Load your custom encoding or the one provided by OpenAI</span> | |
| encoding = get_encoding(<span class="hljs-string">"gpt2"</span>) | |
| convert_tiktoken_to_fast(encoding, <span class="hljs-string">"config/save/dir"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-rdpphz">The resulting <code>tokenizer.json</code> file is saved to the specified directory and loaded with <a href="/docs/transformers/pr_33892/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.from_pretrained">from_pretrained()</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 -->tokenizer = PreTrainedTokenizerFast.from_pretrained(<span class="hljs-string">"config/save/dir"</span>)<!-- HTML_TAG_END --></pre></div> <h2 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></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-p5pbbq">A Transformers model expects the input to be a PyTorch or NumPy tensor. A tokenizer’s job is to preprocess text into those tensors. Specify the framework tensor type to return with the <code>return_tensors</code> parameter.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"google/gemma-2-2b"</span>) | |
| tokenizer(<span class="hljs-string">"We are very happy to show you the 🤗 Transformers library."</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| {<span class="hljs-string">'input_ids'</span>: tensor([[ <span class="hljs-number">2</span>, <span class="hljs-number">1734</span>, <span class="hljs-number">708</span>, <span class="hljs-number">1508</span>, <span class="hljs-number">4915</span>, <span class="hljs-number">577</span>, <span class="hljs-number">1500</span>, <span class="hljs-number">692</span>, <span class="hljs-number">573</span>, | |
| <span class="hljs-number">156808</span>, <span class="hljs-number">128149</span>, <span class="hljs-number">9581</span>, <span class="hljs-number">235265</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>]]) | |
| }<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1bpp0kx">The tokenization process of converting text into input ids is completed in two steps.</p> <div class="flex space-x-2 items-center my-1.5 mr-8 h-7 !pl-0 -mx-3 md:mx-0"><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd border-gray-800 bg-black dark:bg-gray-700 text-white">1. tokenize </div><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd text-gray-500 cursor-pointer opacity-90 hover:text-gray-700 dark:hover:text-gray-200 hover:shadow-sm">2. convert tokens to ids </div><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd text-gray-500 cursor-pointer opacity-90 hover:text-gray-700 dark:hover:text-gray-200 hover:shadow-sm">3. decode ids to text </div></div> <div class="language-select"><p data-svelte-h="svelte-iwbojo">In the first step, a string of text is split into tokens by the <a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizer.tokenize">tokenize()</a> function. How the text is split depends on the tokenization algorithm.</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 -->tokens = tokenizer.tokenize(<span class="hljs-string">"We are very happy to show you the 🤗 Transformers library"</span>) | |
| <span class="hljs-built_in">print</span>(tokens) | |
| [<span class="hljs-string">'We'</span>, <span class="hljs-string">'▁are'</span>, <span class="hljs-string">'▁very'</span>, <span class="hljs-string">'▁happy'</span>, <span class="hljs-string">'▁to'</span>, <span class="hljs-string">'▁show'</span>, <span class="hljs-string">'▁you'</span>, <span class="hljs-string">'▁the'</span>, <span class="hljs-string">'▁🤗'</span>, <span class="hljs-string">'▁Transformers'</span>, <span class="hljs-string">'▁library'</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-mcscva">Gemma uses a <a href="./tokenizer_summary#sentencepiece">SentencePiece</a> tokenizer which replaces spaces with an underscore <code>_</code>.</p> </div> <blockquote class="tip" data-svelte-h="svelte-htiuou"><p>Visualize how different tokenizers work in the <a href="https://xenova-the-tokenizer-playground.static.hf.space" rel="nofollow">Tokenizer Playground</a>.</p></blockquote> <h3 class="relative group"><a id="special-tokens" 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="#special-tokens"><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>Special tokens</span></h3> <p data-svelte-h="svelte-81nneu">Special tokens provide the model with some additional information about the text.</p> <p data-svelte-h="svelte-5q2p53">For example, if you compare the tokens obtained from passing text directly to the tokenizer and from <a href="/docs/transformers/pr_33892/en/main_classes/tokenizer#transformers.PreTrainedTokenizer.convert_tokens_to_ids">convert_tokens_to_ids()</a>, you’ll notice some additional tokens are added.</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 -->model_inputs = tokenizer(<span class="hljs-string">"We are very happy to show you the 🤗 Transformers library."</span>) | |
| [<span class="hljs-number">2</span>, <span class="hljs-number">1734</span>, <span class="hljs-number">708</span>, <span class="hljs-number">1508</span>, <span class="hljs-number">4915</span>, <span class="hljs-number">577</span>, <span class="hljs-number">1500</span>, <span class="hljs-number">692</span>, <span class="hljs-number">573</span>, <span class="hljs-number">156808</span>, <span class="hljs-number">128149</span>, <span class="hljs-number">9581</span>] | |
| tokenizer.convert_tokens_to_ids(tokens) | |
| [<span class="hljs-number">1734</span>, <span class="hljs-number">708</span>, <span class="hljs-number">1508</span>, <span class="hljs-number">4915</span>, <span class="hljs-number">577</span>, <span class="hljs-number">1500</span>, <span class="hljs-number">692</span>, <span class="hljs-number">573</span>, <span class="hljs-number">156808</span>, <span class="hljs-number">128149</span>, <span class="hljs-number">9581</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-2ta9kd">When you <a href="/docs/transformers/pr_33892/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.decode">decode()</a> the ids, you’ll see <code><bos></code> at the beginning of the string. This is used to indicate the beginning of a sentence to the model.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-built_in">print</span>(tokenizer.decode(model_inputs[<span class="hljs-string">"input_ids"</span>])) | |
| <span class="hljs-built_in">print</span>(tokenizer.decode(ids)) | |
| <span class="hljs-string">'<bos>We are very happy to show you the 🤗 Transformers library.'</span> | |
| <span class="hljs-string">'We are very happy to show you the 🤗 Transformers library'</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-10r0jkh">Not all models need special tokens, but if they do, a tokenizer automatically adds them.</p> <h3 class="relative group"><a id="batch-tokenization" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#batch-tokenization"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Batch tokenization</span></h3> <p data-svelte-h="svelte-1pwp3dv">It is faster and more efficient to preprocess <em>batches</em> of text instead of a single sentence at a time. Fast tokenizers are especially good at parallelizing tokenization.</p> <p data-svelte-h="svelte-1502w4">Pass a list of string text to the 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 -->batch_sentences = [ | |
| <span class="hljs-string">"But what about second breakfast?"</span>, | |
| <span class="hljs-string">"Don't think he knows about second breakfast, Pip."</span>, | |
| <span class="hljs-string">"What about elevensies?"</span>, | |
| ] | |
| encoded_inputs = tokenizer(batch_sentences, return_tensors=<span class="hljs-string">"pt"</span>) | |
| <span class="hljs-built_in">print</span>(encoded_inputs) | |
| { | |
| <span class="hljs-string">'input_ids'</span>: | |
| [[<span class="hljs-number">2</span>, <span class="hljs-number">1860</span>, <span class="hljs-number">1212</span>, <span class="hljs-number">1105</span>, <span class="hljs-number">2257</span>, <span class="hljs-number">14457</span>, <span class="hljs-number">235336</span>], | |
| [<span class="hljs-number">2</span>, <span class="hljs-number">4454</span>, <span class="hljs-number">235303</span>, <span class="hljs-number">235251</span>, <span class="hljs-number">1742</span>, <span class="hljs-number">693</span>, <span class="hljs-number">9242</span>, <span class="hljs-number">1105</span>, <span class="hljs-number">2257</span>, <span class="hljs-number">14457</span>, <span class="hljs-number">235269</span>, <span class="hljs-number">48782</span>, <span class="hljs-number">235265</span>], | |
| [<span class="hljs-number">2</span>, <span class="hljs-number">1841</span>, <span class="hljs-number">1105</span>, <span class="hljs-number">29754</span>, <span class="hljs-number">37453</span>, <span class="hljs-number">235336</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>]] | |
| }<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="padding" 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="#padding"><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>Padding</span></h3> <blockquote class="tip" data-svelte-h="svelte-1uztf4f"><p>Learn about additional padding strategies in the <a href="./pad_truncation">Padding and truncation</a> guide.</p></blockquote> <p data-svelte-h="svelte-13g0g3k">In the output above, the <code>input_ids</code> have different lengths. This is an issue because Transformers expects them to have the same lengths so it can pack them into a batch. Sequences with uneven lengths can’t be batched.</p> <p data-svelte-h="svelte-sor0ra">Padding adds a special <em>padding token</em> to ensure all sequences have the same length. Set <code>padding=True</code> to pad the sequences to the longest sequence length in the batch.</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 -->encoded_inputs = tokenizer(batch_sentences, padding=<span class="hljs-literal">True</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| <span class="hljs-built_in">print</span>(encoded_inputs)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-19p5uh3">The tokenizer added the special padding token <code>0</code> to the left side (<em>left padding</em>) because Gemma and LLMs in general are not trained to continue generation from a padding token.</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> <blockquote class="tip" data-svelte-h="svelte-1vpeu21"><p>Learn about additional truncation strategies in the <a href="./pad_truncation">Padding and truncation</a> guide.</p></blockquote> <p data-svelte-h="svelte-pvywvo">Models are only able to process sequences up to a certain length. If you try to process a sequence longer than a model can handle, it crashes.</p> <p data-svelte-h="svelte-x3l9nl">Truncation removes tokens from a sequence to ensure it doesn’t exceed the maximum length. Set <code>truncation=True</code> to truncate a sequence to the maximum length accepted by the model. You can also set the maximum length yourself with the <code>max_length</code> parameter.</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 -->encoded_inputs = tokenizer(batch_sentences, max_length=<span class="hljs-number">8</span>, truncation=<span class="hljs-literal">True</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| <span class="hljs-built_in">print</span>(encoded_inputs)<!-- HTML_TAG_END --></pre></div> <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/en/fast_tokenizers.md" target="_blank"><svg class="mr-1" 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="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg> <span data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a> <p></p> | |
| <script> | |
| { | |
| __sveltekit_16tnnm8 = { | |
| assets: "/docs/transformers/pr_33892/en", | |
| base: "/docs/transformers/pr_33892/en", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/transformers/pr_33892/en/_app/immutable/entry/start.b2c4257a.js"), | |
| import("/docs/transformers/pr_33892/en/_app/immutable/entry/app.05ef1f97.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 23], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 73.5 kB
- Xet hash:
- d06c84f42a31029f27fc89d410bef8584ac32d375596a8693121734b0315586c
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.