Hugging Face's logo

Buckets:
hf-doc-build
/
doc-dev

Files
xet
hf-doc-build/doc-dev / transformers /pr_33913 /it /big_models.html
download
raw
25.9 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Istanziare un big model&quot;,&quot;local&quot;:&quot;istanziare-un-big-model&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Checkpoints condivisi&quot;,&quot;local&quot;:&quot;checkpoints-condivisi&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Caricamento low memory&quot;,&quot;local&quot;:&quot;caricamento-low-memory&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
 <link href="/docs/transformers/pr_33913/it/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/entry/start.cef8f86b.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/scheduler.36a0863c.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/singletons.39cd3b38.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/index.733708bb.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/paths.64312358.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/entry/app.d399d34f.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/index.9c13489a.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/nodes/0.7859fb6f.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/each.e59479a4.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/nodes/6.ca21f5b4.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/Tip.3b06990e.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/CodeBlock.05d8ec32.js">
 <link rel="modulepreload" href="/docs/transformers/pr_33913/it/_app/immutable/chunks/EditOnGithub.e88f2b7b.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Istanziare un big model&quot;,&quot;local&quot;:&quot;istanziare-un-big-model&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Checkpoints condivisi&quot;,&quot;local&quot;:&quot;checkpoints-condivisi&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Caricamento low memory&quot;,&quot;local&quot;:&quot;caricamento-low-memory&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="istanziare-un-big-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="#istanziare-un-big-model"><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>Istanziare un big model</span></h1> <p data-svelte-h="svelte-hpka0t">Quando vuoi utilizzare un modello preaddestrato (pretrained) molto grande, una sfida è minimizzare l’uso della RAM. Il workflow classico
in PyTorch è:</p> <ol data-svelte-h="svelte-1up99oz"><li>Crea il tuo modello con pesi casuali (random weights).</li> <li>Carica i tuoi pesi preaddestrati.</li> <li>Inserisci i pesi preaddestrati nel tuo modello casuale.</li></ol> <p data-svelte-h="svelte-28ptu8">I passi 1 e 2 una versione completa del modello in memoria, in molti casi non è un problema, ma se il modello inizia a pesare diversi GigaBytes, queste due copie possono sturare la nostra RAM. Ancora peggio, se stai usando <code>torch.distributed</code> per seguire l’addestramento (training) in distribuito, ogni processo caricherà il modello preaddestrato e memorizzerà queste due copie nella RAM.</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-exxgkr">Nota che il modello creato casualmente è inizializzato con tensori “vuoti”, che occupano spazio in memoria ma senza riempirlo (quindi i valori casuali sono quelli che si trovavano in questa porzione di memoria in un determinato momento). L’inizializzazione casuale che segue la distribuzione appropriata per il tipo di modello/parametri istanziato (come la distribuzione normale per le istanze) è eseguito solo dopo il passaggio 3 sui pesi non inizializzati, per essere più rapido possibile!</p></div> <p data-svelte-h="svelte-vvoxkl">In questa guida, esploreremo le soluzioni che Transformers offre per affrontare questo problema. C’è da tenere in conto che questa è un’area in cui si sta attualmente sviluppando, quindi le API spiegate qui possono variare velocemente in futuro.</p> <h2 class="relative group"><a id="checkpoints-condivisi" 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="#checkpoints-condivisi"><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>Checkpoints condivisi</span></h2> <p data-svelte-h="svelte-mg2nmg">Dalla versione 4.18.0, i checkpoints dei modelli che occupano più di 10GB di spazio vengono automaticamente frammentati in più parti. Per quanto riguarda la possibilità di avere un unico checkpoint quando si utilizza <code>model.save_pretrained(save_dir)</code>, si hanno diversi checkpoint parziali (ognuno con dimensione &lt; 10GB) e un indice che mappa i nomi dei parametri ai file in cui sono memorizzati.</p> <p data-svelte-h="svelte-13ehqwn">Puoi controllare la dimensione massima dopo la frammentazione con il parametro <code>max_shard_size</code>, nel prossimo esempio, useremo modelli di dimensioni normali con frammenti di piccoli dimensioni: prendiamo un modello BERT classico.</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> AutoModel
model = AutoModel.from_pretrained(<span class="hljs-string">&quot;google-bert/bert-base-cased&quot;</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-pg1prv">Se tu salvi usando <code>save_pretrained()</code>, avrai una nuova cartella con due file: il config del modello e i suoi pesi:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> os
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> tempfile
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">with</span> tempfile.TemporaryDirectory() <span class="hljs-keyword">as</span> tmp_dir:
<span class="hljs-meta">... </span> model.save_pretrained(tmp_dir)
<span class="hljs-meta">... </span> <span class="hljs-built_in">print</span>(<span class="hljs-built_in">sorted</span>(os.listdir(tmp_dir)))
[<span class="hljs-string">&#x27;config.json&#x27;</span>, <span class="hljs-string">&#x27;pytorch_model.bin&#x27;</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1gd3i04">Adesso usiamo una dimensione massima di frammentazione di 200MB:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">with</span> tempfile.TemporaryDirectory() <span class="hljs-keyword">as</span> tmp_dir:
<span class="hljs-meta">... </span> model.save_pretrained(tmp_dir, max_shard_size=<span class="hljs-string">&quot;200MB&quot;</span>)
<span class="hljs-meta">... </span> <span class="hljs-built_in">print</span>(<span class="hljs-built_in">sorted</span>(os.listdir(tmp_dir)))
[<span class="hljs-string">&#x27;config.json&#x27;</span>, <span class="hljs-string">&#x27;pytorch_model-00001-of-00003.bin&#x27;</span>, <span class="hljs-string">&#x27;pytorch_model-00002-of-00003.bin&#x27;</span>, <span class="hljs-string">&#x27;pytorch_model-00003-of-00003.bin&#x27;</span>, <span class="hljs-string">&#x27;pytorch_model.bin.index.json&#x27;</span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-6ew1q4">In aggiunta alla configurazione del modello, vediamo tre differenti file dei pesi, e un file <code>index.json</code> che è il nostro indice. Un checkpoint può essere ricaricato totalmente usando il metodo <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 class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">with</span> tempfile.TemporaryDirectory() <span class="hljs-keyword">as</span> tmp_dir:
<span class="hljs-meta">... </span> model.save_pretrained(tmp_dir, max_shard_size=<span class="hljs-string">&quot;200MB&quot;</span>)
<span class="hljs-meta">... </span> new_model = AutoModel.from_pretrained(tmp_dir)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-5uw3pb">Il vantaggio principale di applicare questo metodo per modelli grandi è che durante il passo 2 del workflow illustrato in precedenza, ogni frammento del checkpoint viene caricato dopo il precedente, limitando l’utilizzo della RAM alla dimensione del modello più la dimensione del frammento più grande.</p> <p data-svelte-h="svelte-7x1jfm">Dietro le quinte, il file indice è utilizzato per determinare quali chiavi sono nel checkpoint, e dove i corrispondenti pesi sono memorizzati. Possiamo caricare l’indice come un qualsiasi json e ottenere un dizionario:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> json
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">with</span> tempfile.TemporaryDirectory() <span class="hljs-keyword">as</span> tmp_dir:
<span class="hljs-meta">... </span> model.save_pretrained(tmp_dir, max_shard_size=<span class="hljs-string">&quot;200MB&quot;</span>)
<span class="hljs-meta">... </span> <span class="hljs-keyword">with</span> <span class="hljs-built_in">open</span>(os.path.join(tmp_dir, <span class="hljs-string">&quot;pytorch_model.bin.index.json&quot;</span>), <span class="hljs-string">&quot;r&quot;</span>) <span class="hljs-keyword">as</span> f:
<span class="hljs-meta">... </span> index = json.load(f)
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-built_in">print</span>(index.keys())
dict_keys([<span class="hljs-string">&#x27;metadata&#x27;</span>, <span class="hljs-string">&#x27;weight_map&#x27;</span>])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-yi9squ">I metadati consistono solo nella dimensione totale del modello per ora. Abbiamo in programma di aggiungere altre informazioni in futuro:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span>index[<span class="hljs-string">&quot;metadata&quot;</span>]
{<span class="hljs-string">&#x27;total_size&#x27;</span>: <span class="hljs-number">433245184</span>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-raulhv">La mappa dei pesi è la parte principale di questo indice, che mappa ogni nome dei parametri (si trova solitamente nei modelli PyTorch come <code>state_dict</code>) al file in cui è memorizzato:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span>index[<span class="hljs-string">&quot;weight_map&quot;</span>]
{<span class="hljs-string">&#x27;embeddings.LayerNorm.bias&#x27;</span>: <span class="hljs-string">&#x27;pytorch_model-00001-of-00003.bin&#x27;</span>,
 <span class="hljs-string">&#x27;embeddings.LayerNorm.weight&#x27;</span>: <span class="hljs-string">&#x27;pytorch_model-00001-of-00003.bin&#x27;</span>,
 ...<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-rh7ba6">Se vuoi caricare direttamente un checkpoint frammentato in un modello senza usare <code>from_pretrained()</code> (come si farebbe con <code>model.load_state_dict()</code> per un checkpoint completo) devi usare <code>load_sharded_checkpoint()</code>:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> transformers.modeling_utils <span class="hljs-keyword">import</span> load_sharded_checkpoint
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">with</span> tempfile.TemporaryDirectory() <span class="hljs-keyword">as</span> tmp_dir:
<span class="hljs-meta">... </span> model.save_pretrained(tmp_dir, max_shard_size=<span class="hljs-string">&quot;200MB&quot;</span>)
<span class="hljs-meta">... </span> load_sharded_checkpoint(model, tmp_dir)<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="caricamento-low-memory" 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="#caricamento-low-memory"><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>Caricamento low memory</span></h2> <p data-svelte-h="svelte-1skddyr">Frammentare i checkpoint l’utilizzo di memoria al passo 2 del workflow citato in precedenza, ma per utilizzare questo modello in un ambiente con poca memoria, consigliamo di utilizzare i nostri strumenti basati sulla libreria Accelerate.</p> <p data-svelte-h="svelte-8rcien">Per ulteriori informazioni, leggere la seguente guida: <a href="./main_classes/model#large-model-loading">Large model loading using Accelerate</a></p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/transformers/blob/main/docs/source/it/big_models.md" target="_blank"><span data-svelte-h="svelte-1kd6by1">&lt;</span> <span data-svelte-h="svelte-x0xyl0">&gt;</span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a> <p></p>
<script>
 {
 __sveltekit_728xau = {
 assets: "/docs/transformers/pr_33913/it",
 base: "/docs/transformers/pr_33913/it",
 env: {}
 };
 const element = document.currentScript.parentElement;
 const data = [null,null];
 Promise.all([
 import("/docs/transformers/pr_33913/it/_app/immutable/entry/start.cef8f86b.js"),
 import("/docs/transformers/pr_33913/it/_app/immutable/entry/app.d399d34f.js")
 ]).then(([kit, app]) => {
 kit.start(app, element, {
 node_ids: [0, 6],
 data,
 form: null,
 error: null
 });
 });
 }
 </script>

Xet Storage Details

Size:
25.9 kB
·
Xet hash:
f04b3fa358e90f64f71678c9d1321e0e53159a64db0ac7fc7e93e7724e761ebe

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