Buckets:

hf-doc-build
/

doc

Files

xet

hf-doc-build/doc / transformers /main /ro /models.html

HuggingFaceDocBuilder

about 3 hours ago

download

raw

46.8 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Încărcarea modelelor","local":"încărcarea-modelelor","sections":[{"title":"Modele și configurații","local":"modele-și-configurații","sections":[],"depth":2},{"title":"Clasele de model","local":"clasele-de-model","sections":[],"depth":2},{"title":"Modele mari","local":"modele-mari","sections":[{"title":"Checkpoint-uri sharded","local":"checkpoint-uri-sharded","sections":[],"depth":3},{"title":"Big Model Inference","local":"big-model-inference","sections":[],"depth":3},{"title":"Tipul de date al modelului","local":"tipul-de-date-al-modelului","sections":[],"depth":3}],"depth":2},{"title":"Modele personalizate","local":"modele-personalizate","sections":[],"depth":2}],"depth":1}"/>
	<link href="/docs/transformers/main/ro/_app/immutable/entry/start.CIgIKwFw.js" rel="modulepreload">
	<link href="/docs/transformers/main/ro/_app/immutable/chunks/BcwjRHJP.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.eNKDC4Vf.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.P_YGbUHv.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/21.Bk9N5Lck.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/BbRxJ3mC.js" rel="modulepreload">
	<link href="/docs/transformers/main/ro/_app/immutable/chunks/hgCuwoId.js" rel="modulepreload">
	<!--18pldtr--><meta name="hf:doc:metadata" content="{"title":"Încărcarea modelelor","local":"încărcarea-modelelor","sections":[{"title":"Modele și configurații","local":"modele-și-configurații","sections":[],"depth":2},{"title":"Clasele de model","local":"clasele-de-model","sections":[],"depth":2},{"title":"Modele mari","local":"modele-mari","sections":[{"title":"Checkpoint-uri sharded","local":"checkpoint-uri-sharded","sections":[],"depth":3},{"title":"Big Model Inference","local":"big-model-inference","sections":[],"depth":3},{"title":"Tipul de date al modelului","local":"tipul-de-date-al-modelului","sections":[],"depth":3}],"depth":2},{"title":"Modele personalizate","local":"modele-personalizate","sections":[],"depth":2}],"depth":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="încărcarea-modelelor" 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="#încărcarea-modelelor"><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>Încărcarea modelelor</span></h1><!--]--><!----> <p>Transformers oferă multe modele pre-antrenate gata de utilizare cu o singură linie de cod. Acestea necesită o clasă de model și metoda <code>from_pretrained()</code>.</p> <p>Apelează <code>from_pretrained()</code> pentru a descărca și încărca weights și configurația unui model stocate pe <a href="https://hf.co/models" rel="nofollow">Hub-ul Hugging Face</a>.</p> <blockquote class="tip"><p>Metoda <code>from_pretrained()</code> încarcă weights stocate în formatul de fișier <a href="https://hf.co/docs/safetensors/index" rel="nofollow">safetensors</a> dacă sunt disponibile. În mod tradițional, weights modelelor PyTorch sunt serializate cu utilitarul <a href="https://docs.python.org/3/library/pickle.html" rel="nofollow">pickle</a> care este cunoscut ca nesecurizat. Fișierele safetensors sunt mai sigure și mai rapid de încărcat.</p></blockquote> <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> AutoModelForCausalLM

	model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"meta-llama/Llama-2-7b-hf"</span>, device_map=<span class="hljs-string">"auto"</span>)<!----></pre></div><!----> <p>Acest ghid explică cum sunt încărcate modelele, diferitele moduri de a încărca un model, cum să depășești problemele de memorie pentru modele foarte mari și cum să încarci modele personalizate.</p> <!--[1--><h2 class="relative group"><a id="modele-și-configurații" 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="#modele-și-configurații"><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>Modele și configurații</span></h2><!--]--><!----> <p>Toate modelele au un fișier <code>configuration.py</code> cu atribute specifice precum numărul de hidden layers, dimensiunea vocabularului, funcția de activare și altele. Vei găsi și un fișier <code>modeling.py</code> care definește layers și operațiile matematice din interiorul fiecărui strat. Fișierul <code>modeling.py</code> preia atributele modelului din <code>configuration.py</code> și construiește modelul corespunzător. În acest moment, ai un model cu weights aleatorii care trebuie antrenat pentru a produce rezultate semnificative.</p> <blockquote class="tip"><p>O <em>arhitectură</em> se referă la scheletul modelului, iar un <em>checkpoint</em> se referă la weights pentru o anumită arhitectură. De exemplu, [BERT] este o arhitectură, în timp ce <a href="https://huggingface.co/google-bert/bert-base-uncased" rel="nofollow">google-bert/bert-base-uncased</a> este un checkpoint. Vei vedea termenul <em>model</em> utilizat interschimbabil cu arhitectura și checkpoint-ul.</p></blockquote> <p>Există două tipuri generale de modele pe care le poți încărca:</p> <ol><li>Un model de bază, precum <code>AutoModel</code> sau <code>LlamaModel</code>, care returnează hidden states.</li> <li>Un model cu un <em>head</em> specific atașat, precum <code>AutoModelForCausalLM</code> sau <code>LlamaForCausalLM</code>, pentru efectuarea unor task-uri specifice.</li></ol> <!--[1--><h2 class="relative group"><a id="clasele-de-model" 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-model"><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 model</span></h2><!--]--><!----> <p>Pentru a obține un model pre-antrenat, trebuie să încarci weights în model. Acest lucru se face apelând <code>from_pretrained()</code> care acceptă weights de pe Hub-ul Hugging Face sau dintr-un folder local.</p> <p>Există două clase de model: clasa [AutoModel] și o clasă 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">AutoModel</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 class</div><!--]--></div> <div class="language-select"><!--[0--><iframe class="w-full xl:w-4/6 h-80" src="https://www.youtube-nocookie.com/embed/AhChOFRegn4" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen=""></iframe><!----> <p>Clasa [AutoModel] este o modalitate convenabilă de a încărca o arhitectură fără a fi nevoie să cunoști numele exact al clasei de model, deoarece există multe modele disponibile. Selectează automat clasa de model corectă pe baza fișierului de configurație. Trebuie să știi doar task-ul și checkpoint-ul pe care vrei să le utilizezi.</p> <p>Comută ușor între modele sau task-uri, atât timp cât arhitectura este suportată pentru un task dat.</p> <p>De exemplu, același model poate fi utilizat pentru task-uri separate.</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> AutoModelForCausalLM, AutoModelForSequenceClassification, AutoModelForQuestionAnswering

	<span class="hljs-comment"># utilizează același API pentru 3 task-uri diferite</span>
	model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"meta-llama/Llama-2-7b-hf"</span>)
	model = AutoModelForSequenceClassification.from_pretrained(<span class="hljs-string">"meta-llama/Llama-2-7b-hf"</span>)
	model = AutoModelForQuestionAnswering.from_pretrained(<span class="hljs-string">"meta-llama/Llama-2-7b-hf"</span>)<!----></pre></div><!----> <p>În alte cazuri, ai putea testa rapid mai multe modele diferite pentru un task.</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> AutoModelForCausalLM

	<span class="hljs-comment"># utilizează același API pentru 3 modele diferite</span>
	model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"meta-llama/Llama-2-7b-hf"</span>)
	model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"mistralai/Mistral-7B-v0.1"</span>)
	model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"google/gemma-7b"</span>)<!----></pre></div><!----><!----><!--]--><!----> <!--[-1--><!--]--><!----><!----></div><!----> <!--[1--><h2 class="relative group"><a id="modele-mari" 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="#modele-mari"><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>Modele mari</span></h2><!--]--><!----> <p>Modelele mari pre-antrenate necesită multă memorie pentru a fi încărcate. Procesul de încărcare implică:</p> <ol><li>crearea unui model cu weights aleatoare</li> <li>încărcarea weights pre-antrenate</li> <li>plasarea weights pre-antrenate pe model</li></ol> <p>Ai nevoie de suficientă memorie pentru a reține două copii ale weights modelului (aleatoare și pre-antrenate), ceea ce poate să nu fie posibil în funcție de hardware-ul tău. În medii de antrenare distribuită, aceasta este și mai dificilă deoarece fiecare proces încarcă un model pre-antrenat.</p> <p>Transformers reduce unele dintre aceste provocări legate de memorie prin inițializare rapidă, checkpoint-uri sharded, funcția <a href="https://hf.co/docs/accelerate/usage_guides/big_modeling" rel="nofollow">Big Model Inference</a> din Accelerate și suportul pentru tipuri de date cu mai puțini biți.</p> <!--[2--><h3 class="relative group"><a id="checkpoint-uri-sharded" 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="#checkpoint-uri-sharded"><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>Checkpoint-uri sharded</span></h3><!--]--><!----> <p><code>save_pretrained()</code> shardează automat checkpoint-urile mai mari de 50GB. Aceasta menține numărul de shards redus pentru modelele mari și simplifică gestionarea fișierelor.</p> <p>Parametrii se încarcă în paralel și vârful de memorie depinde doar de dimensiunea modelului. Utilizează <code>max_shard_size</code> în <code>save_pretrained()</code> pentru a seta dimensiunea maximă a checkpoint-ului înainte de sharding.</p> <blockquote class="note"><p>Utilizarea memoriei pentru modelele care necesită conversie dinamică de weights depinde de dimensiunea modelului și de dimensiunea celor mai mari parametri dintr-o singură conversie. Aceasta se aplică de obicei modelelor mixture-of-experts (MoE) unde utilizarea memoriei este dimensiunea modelului plus numărul de experți de pe un strat. Consultă ghidul <a href="./weightconverter#%C3%AEnc%C4%83rcarea-rapid%C4%83-%C8%99i-eficient%C4%83-a-modelelor">încărcarea dinamică a modelelor</a> pentru a afla mai multe despre cum sunt încărcate modelele.</p></blockquote> <p><code>save_pretrained()</code> creează și un fișier index care mapează numele parametrilor la fișierele lor de shard. Index-ul conține două chei, <code>metadata</code> și <code>weight_map</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">import</span> json

	<span class="hljs-keyword">with</span> tempfile.TemporaryDirectory() <span class="hljs-keyword">as</span> tmp_dir:
	model.save_pretrained(tmp_dir, max_shard_size=<span class="hljs-string">"50GB"</span>)
	<span class="hljs-keyword">with</span> <span class="hljs-built_in">open</span>(os.path.join(tmp_dir, <span class="hljs-string">"model.safetensors.index.json"</span>), <span class="hljs-string">"r"</span>) <span class="hljs-keyword">as</span> f:
	index = json.load(f)

	<span class="hljs-built_in">print</span>(index.keys())<!----></pre></div><!----> <p><code>metadata</code> stochează dimensiunea totală a modelului.</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 "><!---->index[<span class="hljs-string">"metadata"</span>]
	{<span class="hljs-string">'total_size'</span>: <span class="hljs-number">28966928384</span>}<!----></pre></div><!----> <p><code>weight_map</code> mapează fiecare parametru la fișierul său de shard.</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 "><!---->index[<span class="hljs-string">"weight_map"</span>]
	{<span class="hljs-string">'lm_head.weight'</span>: <span class="hljs-string">'model-00006-of-00006.safetensors'</span>,
	<span class="hljs-string">'model.embed_tokens.weight'</span>: <span class="hljs-string">'model-00001-of-00006.safetensors'</span>,
	<span class="hljs-string">'model.layers.0.input_layernorm.weight'</span>: <span class="hljs-string">'model-00001-of-00006.safetensors'</span>,
	<span class="hljs-string">'model.layers.0.mlp.down_proj.weight'</span>: <span class="hljs-string">'model-00001-of-00006.safetensors'</span>,
	...
	}<!----></pre></div><!----> <!--[2--><h3 class="relative group"><a id="big-model-inference" 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="#big-model-inference"><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>Big Model Inference</span></h3><!--]--><!----> <blockquote class="tip"><p>Asigură-te că ai Accelerate v0.9.0 și PyTorch v1.9.0 sau o versiune mai nouă instalate pentru a utiliza această funcție!</p></blockquote> <iframe class="w-full xl:w-4/6 h-80" src="https://www.youtube-nocookie.com/embed/MWCSGj9jEAo" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen=""></iframe><!----> <p><code>from_pretrained()</code> este potențat cu funcția <a href="https://hf.co/docs/accelerate/usage_guides/big_modeling" rel="nofollow">Big Model Inference</a> din Accelerate.</p> <p>Big Model Inference creează un <em>schelet de model</em> pe dispozitivul <a href="https://pytorch.org/docs/main/meta.html" rel="nofollow">meta</a> PyTorch. Dispozitivul meta nu stochează date reale, doar metadata.</p> <p>Weights inițializate aleatoriu sunt create doar când sunt încărcate weights pre-antrenate, pentru a evita menținerea în memorie a două copii ale modelului în același timp. Utilizarea maximă a memoriei este doar dimensiunea modelului.</p> <blockquote class="tip"><p>Află mai multe despre plasarea pe dispozitive în <a href="https://hf.co/docs/accelerate/v0.33.0/en/concept_guides/big_model_inference#designing-a-device-map" rel="nofollow">Designing a device map</a>.</p></blockquote> <p>A doua funcție a Big Model Inference se referă la modul în care weights sunt încărcate și distribuite în scheletul modelului. Weights modelului sunt distribuite pe toate dispozitivele disponibile, începând cu cel mai rapid dispozitiv (de obicei GPU) și descărcând ulterior weights rămase pe dispozitive mai lente (CPU și hard disk).</p> <p>Ambele funcții combinate reduc utilizarea memoriei și timpii de încărcare pentru modelele mari pre-antrenate.</p> <p>Setează <a href="https://github.com/huggingface/transformers/blob/026a173a64372e9602a16523b8fae9de4b0ff428/src/transformers/modeling_utils.py#L3061" rel="nofollow">device_map</a> la <code>"auto"</code> pentru a activa Big Model Inference.</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> AutoModelForCausalLM

	model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"google/gemma-7b"</span>, device_map=<span class="hljs-string">"auto"</span>)<!----></pre></div><!----> <p>Poți, de asemenea, să atribui manual straturi unui dispozitiv în <code>device_map</code>. Ar trebui să mapeze toți parametrii modelului la un dispozitiv, dar nu trebuie să detaliezi unde merg toate sub-modulele unui strat dacă întregul strat se află pe același dispozitiv.</p> <p>Accesează atributul <code>hf_device_map</code> pentru a vedea cum este distribuit un model pe dispozitive.</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 "><!---->device_map = {<span class="hljs-string">"model.layers.1"</span>: <span class="hljs-number">0</span>, <span class="hljs-string">"model.layers.14"</span>: <span class="hljs-number">1</span>, <span class="hljs-string">"model.layers.31"</span>: <span class="hljs-string">"cpu"</span>, <span class="hljs-string">"lm_head"</span>: <span class="hljs-string">"disk"</span>}
	model.hf_device_map<!----></pre></div><!----> <!--[2--><h3 class="relative group"><a id="tipul-de-date-al-modelului" 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="#tipul-de-date-al-modelului"><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>Tipul de date al modelului</span></h3><!--]--><!----> <p>Argumentul <code>dtype</code> controlează <a href="https://pytorch.org/docs/stable/tensor_attributes.html#torch.dtype" rel="nofollow">dtype-ul</a> PyTorch utilizat pentru a instanția weights modelului. În mod implicit, Transformers încarcă weights cu valoarea <code>dtype</code> sau a moștenitorului <code>torch_dtype</code> din <code>config.json</code>. Dacă <code>config.json</code> nu include niciuna dintre aceste valori, Transformers utilizează dtype-ul primului weight în virgulă mobilă din checkpoint.</p> <p>Suprascrie valoarea implicită pasând un dtype specific.</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">import</span> torch
	<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoModelForCausalLM

	<span class="hljs-comment"># specific dtype</span>
	model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"google/gemma-3-1b-it"</span>, dtype=torch.float16)<!----></pre></div><!----> <p><code>AutoConfig</code> acceptă de asemenea <code>dtype</code> pentru modelele instanțiate de la zero.</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">import</span> torch
	<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoConfig, AutoModel

	my_config = AutoConfig.from_pretrained(<span class="hljs-string">"google/gemma-2b"</span>, dtype=torch.float16)
	model = AutoModel.from_config(my_config)<!----></pre></div><!----> <!--[1--><h2 class="relative group"><a id="modele-personalizate" 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="#modele-personalizate"><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>Modele personalizate</span></h2><!--]--><!----> <p>Modelele personalizate se construiesc pe clasele de configurație și modelare ale Transformers, suportă API-ul <a href="#autoclass">AutoClass</a> și sunt încărcate cu <code>from_pretrained()</code>. Diferența este că codul de modelare <em>nu</em> provine din Transformers.</p> <p>Fii atent la încărcarea unui model personalizat. Deși Hub-ul include <a href="https://hf.co/docs/hub/security-malware#malware-scanning" rel="nofollow">scanare de malware</a> pentru fiecare repository, trebuie totuși să fii atent să nu execuți accidental cod malițios.</p> <p>Setează <code>trust_remote_code=True</code> în <code>from_pretrained()</code> pentru a încărca un model personalizat.</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> AutoModelForImageClassification

	model = AutoModelForImageClassification.from_pretrained(<span class="hljs-string">"sgugger/custom-resnet50d"</span>, trust_remote_code=<span class="hljs-literal">True</span>)<!----></pre></div><!----> <p>Ca un nivel suplimentar de securitate, încarcă un model personalizat dintr-o revizuire specifică pentru a evita încărcarea codului de model care ar putea fi schimbat. Hash-ul commit-ului poate fi copiat din <a href="https://hf.co/sgugger/custom-resnet50d/commits/main" rel="nofollow">istoricul commit-urilor</a> modelului.</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 "><!---->commit_hash = <span class="hljs-string">"ed94a7c6247d8aedce4647f00f20de6875b5b292"</span>
	model = AutoModelForImageClassification.from_pretrained(
	<span class="hljs-string">"sgugger/custom-resnet50d"</span>, trust_remote_code=<span class="hljs-literal">True</span>, revision=commit_hash
	)<!----></pre></div><!----> <p>Consultă ghidul <a href="./custom_models">Personalizarea modelelor</a> pentru mai multe informații.</p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/transformers/blob/main/docs/source/ro/models.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_1pz8pps = {
	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.CIgIKwFw.js"),
	import("/docs/transformers/main/ro/_app/immutable/entry/app.eNKDC4Vf.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 21],
	data: [null,null],
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 46.8 kB
Xet hash:: 7f1e3cb8198a4116044f500fa3a80c790cec58d517d215d1ddec2c0ce6165ba2

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