Buckets:

hf-doc-build/doc / transformers /main /ro /feature_extractors.html
HuggingFaceDocBuilder's picture
download
raw
39.8 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Feature extractors&quot;,&quot;local&quot;:&quot;feature-extractors&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Clasele de feature extractor&quot;,&quot;local&quot;:&quot;clasele-de-feature-extractor&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Preprocesare&quot;,&quot;local&quot;:&quot;preprocesare&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Padding&quot;,&quot;local&quot;:&quot;padding&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Trunchiere&quot;,&quot;local&quot;:&quot;trunchiere&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Resampling&quot;,&quot;local&quot;:&quot;resampling&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/>
<link href="/docs/transformers/main/ro/_app/immutable/entry/start.DwnOcLC_.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/CL87Xr5_.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/Cl6kX2N0.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/entry/app.Dw7RcaNl.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/BzzeShVL.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/DWwSmaYb.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/DM0RASQS.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/DsnmJJEf.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/nodes/0.DMmsXSdS.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/C8aKV6he.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/nodes/12.BeC4bqEg.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/CjTxHwdC.js" rel="modulepreload">
<link href="/docs/transformers/main/ro/_app/immutable/chunks/hgCuwoId.js" rel="modulepreload">
<!--mfpjxz--><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Feature extractors&quot;,&quot;local&quot;:&quot;feature-extractors&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Clasele de feature extractor&quot;,&quot;local&quot;:&quot;clasele-de-feature-extractor&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Preprocesare&quot;,&quot;local&quot;:&quot;preprocesare&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Padding&quot;,&quot;local&quot;:&quot;padding&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Trunchiere&quot;,&quot;local&quot;:&quot;trunchiere&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;Resampling&quot;,&quot;local&quot;:&quot;resampling&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2}],&quot;depth&quot;:1}"/><!---->
<link href="/docs/transformers/main/ro/_app/immutable/assets/0.tn0RQdqM.css" rel="modulepreload"> <!--[--><!--[0--><!--[--><!--[0--><!--[--><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 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm 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 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" 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-7 max-sm:h-7 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 sm:size-3.5 size-3 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> <!--[-1--><!--]--></div><!----> <!--[0--><h1 class="relative group"><a id="feature-extractors" 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-extractors"><span><svg 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 extractors</span></h1><!--]--><!----> <p>Feature extractors preprocesează datele audio în formatul corect pentru un model dat. Preia semnalul audio brut și îl convertește într-un tensor care poate fi pasat unui model. Forma tensorului depinde de model, dar feature extractor-ul va preprocesa corect datele audio pentru modelul pe care îl folosești. Feature extractors includ și metode pentru padding, trunchiere și resampling.</p> <p>Apelează <code>from_pretrained()</code> ca să încarci un feature extractor și configurația sa de preprocesare de pe Hub-ul Hugging Face sau dintr-un director local. Configurația feature extractor-ului și a preprocesatorului este salvată într-un fișier <a href="https://hf.co/openai/whisper-tiny/blob/main/preprocessor_config.json" rel="nofollow">preprocessor_config.json</a>.</p> <p>Pasează semnalul audio, stocat de obicei în <code>array</code>, feature extractor-ului și setează parametrul <code>sampling_rate</code> la rata de eșantionare a modelului audio pre-antrenat. Este important ca rata de eșantionare a datelor audio să se potrivească cu rata de eșantionare a datelor pe care a fost antrenat modelul audio pre-antrenat.</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 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="language-py "><!----><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoFeatureExtractor
feature_extractor = AutoFeatureExtractor.from_pretrained(<span class="hljs-string">&quot;facebook/wav2vec2-base&quot;</span>)
dataset = load_dataset(<span class="hljs-string">&quot;PolyAI/minds14&quot;</span>, name=<span class="hljs-string">&quot;en-US&quot;</span>, split=<span class="hljs-string">&quot;train&quot;</span>)
processed_sample = feature_extractor(dataset[<span class="hljs-number">0</span>][<span class="hljs-string">&quot;audio&quot;</span>][<span class="hljs-string">&quot;array&quot;</span>], sampling_rate=<span class="hljs-number">16000</span>)
processed_sample
{<span class="hljs-string">&#x27;input_values&#x27;</span>: [array([ <span class="hljs-number">9.4472744e-05</span>, <span class="hljs-number">3.0777880e-03</span>, -<span class="hljs-number">2.8888427e-03</span>, ...,
-<span class="hljs-number">2.8888427e-03</span>, <span class="hljs-number">9.4472744e-05</span>, <span class="hljs-number">9.4472744e-05</span>], dtype=float32)]}<!----></pre></div><!----> <p>Feature extractor-ul returnează un input, <code>input_values</code>, gata să fie consumat de model.</p> <p>Acest ghid te ghidează prin clasele de feature extractor și cum să preprocesezi date audio.</p> <!--[1--><h2 class="relative group"><a id="clasele-de-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="#clasele-de-feature-extractor"><span><svg 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>Clasele de feature extractor</span></h2><!--]--><!----> <p>Feature extractors din Transformers moștenesc din clasa de bază <code>SequenceFeatureExtractor</code> care subclasează <code>FeatureExtractionMixin</code>.</p> <ul><li><code>SequenceFeatureExtractor</code> furnizează o metodă de <code>pad()</code> pentru secvențe la o anumită lungime ca să eviți lungimile inegale ale secvențelor.</li> <li><code>FeatureExtractionMixin</code> furnizează <code>from_pretrained()</code> și <code>save_pretrained()</code> ca să încarci și să salvezi un feature extractor.</li></ul> <p>Există două moduri de a încărca un feature extractor: <code>AutoFeatureExtractor</code> și o clasă de feature extractor specifică modelului.</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">AutoFeatureExtractor</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 feature extractor</div><!--]--></div> <div class="language-select"><!--[0--><p>API-ul [AutoClass] încarcă automat feature extractor-ul corect pentru un model dat.</p> <p>Folosește <code>from_pretrained()</code> ca să încarci un feature extractor.</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 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="language-py "><!----><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoFeatureExtractor
feature_extractor = AutoFeatureExtractor.from_pretrained(<span class="hljs-string">&quot;openai/whisper-tiny&quot;</span>)<!----></pre></div><!----><!----><!--]--><!----> <!--[-1--><!--]--><!----><!----></div><!----> <!--[1--><h2 class="relative group"><a id="preprocesare" 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="#preprocesare"><span><svg 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>Preprocesare</span></h2><!--]--><!----> <p>Un feature extractor așteaptă un tensor PyTorch de o anumită formă ca input. Forma exactă a input-ului poate varia în funcție de modelul audio specific pe care îl folosești.</p> <p>De exemplu, <a href="https://huggingface.co/docs/transformers/model_doc/whisper" rel="nofollow">Whisper</a> se așteaptă ca <code>input_features</code> să fie un tensor de forma <code>(batch_size, feature_size, sequence_length)</code>, dar <a href="https://hf.co/docs/transformers/model_doc/wav2vec2" rel="nofollow">Wav2Vec2</a> se așteaptă ca <code>input_values</code> să fie un tensor de forma <code>(batch_size, sequence_length)</code>.</p> <p>Feature extractor-ul generează forma corectă de input pentru orice model audio folosești.</p> <p>Un feature extractor setează și rata de eșantionare (numărul de valori ale semnalului audio preluate pe secundă) ale fișierelor audio. Rata de eșantionare a datelor tale audio trebuie să se potrivească cu rata de eșantionare a dataset-ului pe care a fost antrenat un model preantrenat. Această valoare este dată de obicei în card-ul modelului.</p> <p>Încarcă un dataset și un feature extractor cu <code>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 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="language-py "><!----><span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset, Audio
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoFeatureExtractor
dataset = load_dataset(<span class="hljs-string">&quot;PolyAI/minds14&quot;</span>, name=<span class="hljs-string">&quot;en-US&quot;</span>, split=<span class="hljs-string">&quot;train&quot;</span>)
feature_extractor = AutoFeatureExtractor.from_pretrained(<span class="hljs-string">&quot;facebook/wav2vec2-base&quot;</span>)<!----></pre></div><!----> <p>Uită-te la primul exemplu din dataset și accesează coloana <code>audio</code> care conține <code>array</code>, semnalul audio brut.</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 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="language-py "><!---->dataset[<span class="hljs-number">0</span>][<span class="hljs-string">&quot;audio&quot;</span>][<span class="hljs-string">&quot;array&quot;</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> ])<!----></pre></div><!----> <p>Feature extractor-ul preprocesează <code>array</code> în formatul de input așteptat pentru un model audio dat. Folosește parametrul <code>sampling_rate</code> ca să setezi rata de eșantionare corespunzătoare.</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 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="language-py "><!---->processed_dataset = feature_extractor(dataset[<span class="hljs-number">0</span>][<span class="hljs-string">&quot;audio&quot;</span>][<span class="hljs-string">&quot;array&quot;</span>], sampling_rate=<span class="hljs-number">16000</span>)
processed_dataset
{<span class="hljs-string">&#x27;input_values&#x27;</span>: [array([ <span class="hljs-number">9.4472744e-05</span>, <span class="hljs-number">3.0777880e-03</span>, -<span class="hljs-number">2.8888427e-03</span>, ...,
-<span class="hljs-number">2.8888427e-03</span>, <span class="hljs-number">9.4472744e-05</span>, <span class="hljs-number">9.4472744e-05</span>], dtype=float32)]}<!----></pre></div><!----> <!--[2--><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 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><!--]--><!----> <p>Lungimile diferite ale secvențelor audio sunt o problemă pentru că Transformers se așteaptă ca toate secvențele să aibă aceeași lungime ca să fie grupate în batch-uri. Secvențele de lungimi inegale nu pot fi grupate în batch-uri.</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 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="language-py "><!---->dataset[<span class="hljs-number">0</span>][<span class="hljs-string">&quot;audio&quot;</span>][<span class="hljs-string">&quot;array&quot;</span>].shape
(<span class="hljs-number">86699</span>,)
dataset[<span class="hljs-number">1</span>][<span class="hljs-string">&quot;audio&quot;</span>][<span class="hljs-string">&quot;array&quot;</span>].shape
(<span class="hljs-number">53248</span>,)<!----></pre></div><!----> <p>Padding-ul adaugă un <em>token de padding</em> special ca să se asigure că toate secvențele au aceeași lungime. Feature extractor-ul adaugă un <code>0</code> — interpretat ca tăcere — la <code>array</code> ca să facă padding. Setează <code>padding=True</code> ca să faci padding secvențelor la lungimea celei mai lungi secvențe din 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 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="language-py "><!----><span class="hljs-keyword">def</span> <span class="hljs-title function_">preprocess_function</span>(<span class="hljs-params">examples</span>):
audio_arrays = [x[<span class="hljs-string">&quot;array&quot;</span>] <span class="hljs-keyword">for</span> x <span class="hljs-keyword">in</span> examples[<span class="hljs-string">&quot;audio&quot;</span>]]
inputs = feature_extractor(
audio_arrays,
sampling_rate=<span class="hljs-number">16000</span>,
padding=<span class="hljs-literal">True</span>,
)
<span class="hljs-keyword">return</span> inputs
processed_dataset = preprocess_function(dataset[:<span class="hljs-number">5</span>])
processed_dataset[<span class="hljs-string">&quot;input_values&quot;</span>][<span class="hljs-number">0</span>].shape
(<span class="hljs-number">86699</span>,)
processed_dataset[<span class="hljs-string">&quot;input_values&quot;</span>][<span class="hljs-number">1</span>].shape
(<span class="hljs-number">86699</span>,)<!----></pre></div><!----> <!--[2--><h3 class="relative group"><a id="trunchiere" 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="#trunchiere"><span><svg 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>Trunchiere</span></h3><!--]--><!----> <p>Modelele pot procesa secvențe doar până la o anumită lungime înainte să se blocheze.</p> <p>Truncherea este o strategie de eliminare a token-urilor în exces dintr-o secvență ca să se asigure că nu depășește lungimea maximă. Setează <code>truncation=True</code> ca să trunchiezi o secvență la lungimea din parametrul <code>max_length</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 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="language-py "><!----><span class="hljs-keyword">def</span> <span class="hljs-title function_">preprocess_function</span>(<span class="hljs-params">examples</span>):
audio_arrays = [x[<span class="hljs-string">&quot;array&quot;</span>] <span class="hljs-keyword">for</span> x <span class="hljs-keyword">in</span> examples[<span class="hljs-string">&quot;audio&quot;</span>]]
inputs = feature_extractor(
audio_arrays,
sampling_rate=<span class="hljs-number">16000</span>,
max_length=<span class="hljs-number">50000</span>,
truncation=<span class="hljs-literal">True</span>,
)
<span class="hljs-keyword">return</span> inputs
processed_dataset = preprocess_function(dataset[:<span class="hljs-number">5</span>])
processed_dataset[<span class="hljs-string">&quot;input_values&quot;</span>][<span class="hljs-number">0</span>].shape
(<span class="hljs-number">50000</span>,)
processed_dataset[<span class="hljs-string">&quot;input_values&quot;</span>][<span class="hljs-number">1</span>].shape
(<span class="hljs-number">50000</span>,)<!----></pre></div><!----> <!--[2--><h3 class="relative group"><a id="resampling" 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="#resampling"><span><svg 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>Resampling</span></h3><!--]--><!----> <p>Biblioteca <a href="https://hf.co/docs/datasets/index" rel="nofollow">Datasets</a> poate și să resample datele audio ca să se potrivească cu rata de eșantionare așteptată de un model audio. Metoda asta resamplează datele audio din mers când sunt încărcate, ceea ce poate fi mai rapid decât resamplingul întregului dataset în loc.</p> <p>Dataset-ul audio cu care lucrezi are o rată de eșantionare de 8kHz, iar modelul preantrenat se așteaptă la 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 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="language-py "><!---->dataset[<span class="hljs-number">0</span>][<span class="hljs-string">&quot;audio&quot;</span>]
{<span class="hljs-string">&#x27;path&#x27;</span>: <span class="hljs-string">&#x27;/root/.cache/huggingface/datasets/downloads/extracted/f507fdca7f475d961f5bb7093bcc9d544f16f8cab8608e772a2ed4fbeb4d6f50/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav&#x27;</span>,
<span class="hljs-string">&#x27;array&#x27;</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> ]),
<span class="hljs-string">&#x27;sampling_rate&#x27;</span>: <span class="hljs-number">8000</span>}<!----></pre></div><!----> <p>Apelează <code>cast_column</code> pe coloana <code>audio</code> ca să mărești rata de eșantionare la 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 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="language-py "><!---->dataset = dataset.cast_column(<span class="hljs-string">&quot;audio&quot;</span>, Audio(sampling_rate=<span class="hljs-number">16000</span>))<!----></pre></div><!----> <p>Când încarci sample-ul din dataset, acum este resampled la 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 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="language-py "><!---->dataset[<span class="hljs-number">0</span>][<span class="hljs-string">&quot;audio&quot;</span>]
{<span class="hljs-string">&#x27;path&#x27;</span>: <span class="hljs-string">&#x27;/root/.cache/huggingface/datasets/downloads/extracted/f507fdca7f475d961f5bb7093bcc9d544f16f8cab8608e772a2ed4fbeb4d6f50/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav&#x27;</span>,
<span class="hljs-string">&#x27;array&#x27;</span>: array([ <span class="hljs-number">1.70562416e-05</span>, <span class="hljs-number">2.18727451e-04</span>, <span class="hljs-number">2.28099874e-04</span>, ...,
<span class="hljs-number">3.43842403e-05</span>, -<span class="hljs-number">5.96364771e-06</span>, -<span class="hljs-number">1.76846661e-05</span>]),
<span class="hljs-string">&#x27;sampling_rate&#x27;</span>: <span class="hljs-number">16000</span>}<!----></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/ro/feature_extractors.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><span class="underline">Update</span> on GitHub</span></a><!----> <p></p><!--]--><!----><!--]--><!--]--><!--]--> <!--[-1--><!--]--><!--]-->
<script>
{
__sveltekit_10j3czc = {
base: "/docs/transformers/main/ro",
assets: "/docs/transformers/main/ro"
};
const element = document.currentScript.parentElement;
Promise.all([
import("/docs/transformers/main/ro/_app/immutable/entry/start.DwnOcLC_.js"),
import("/docs/transformers/main/ro/_app/immutable/entry/app.Dw7RcaNl.js")
]).then(([kit, app]) => {
kit.start(app, element, {
node_ids: [0, 12],
data: [null,null],
form: null,
error: null
});
});
}
</script>

Xet Storage Details

Size:
39.8 kB
·
Xet hash:
5849e7c7dda9c0cb34cab04639dbcd325aa9538815a7a47f8484a00b992e9353

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