Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Condividere modelli personalizzati","local":"condividere-modelli-personalizzati","sections":[{"title":"Scrivere una configurazione personalizzata","local":"scrivere-una-configurazione-personalizzata","sections":[],"depth":2},{"title":"Scrivere un modello personalizzato","local":"scrivere-un-modello-personalizzato","sections":[],"depth":2},{"title":"Inviare il codice all’Hub","local":"inviare-il-codice-allhub","sections":[],"depth":2},{"title":"Usare un modello con codice personalizzato","local":"usare-un-modello-con-codice-personalizzato","sections":[],"depth":2},{"title":"Registrare un modello con codice personalizzato nelle classi Auto","local":"registrare-un-modello-con-codice-personalizzato-nelle-classi-auto","sections":[],"depth":2}],"depth":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/10.8dcbf384.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="{"title":"Condividere modelli personalizzati","local":"condividere-modelli-personalizzati","sections":[{"title":"Scrivere una configurazione personalizzata","local":"scrivere-una-configurazione-personalizzata","sections":[],"depth":2},{"title":"Scrivere un modello personalizzato","local":"scrivere-un-modello-personalizzato","sections":[],"depth":2},{"title":"Inviare il codice all’Hub","local":"inviare-il-codice-allhub","sections":[],"depth":2},{"title":"Usare un modello con codice personalizzato","local":"usare-un-modello-con-codice-personalizzato","sections":[],"depth":2},{"title":"Registrare un modello con codice personalizzato nelle classi Auto","local":"registrare-un-modello-con-codice-personalizzato-nelle-classi-auto","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="condividere-modelli-personalizzati" 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="#condividere-modelli-personalizzati"><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>Condividere modelli personalizzati</span></h1> <p data-svelte-h="svelte-xqfdfm">La libreria 🤗 Transformers è studiata per essere facilmente estendibile. Il codice di ogni modello è interamente | |
| situato in una sottocartella del repository senza alcuna astrazione, perciò puoi facilmente copiare il file di un | |
| modello e modificarlo in base ai tuoi bisogni.</p> <p data-svelte-h="svelte-9vqb6y">Se stai scrivendo un nuovo modello, potrebbe essere più semplice iniziare da zero. In questo tutorial, ti mostreremo | |
| come scrivere un modello personalizzato e la sua configurazione in modo che possa essere utilizzato all’interno di | |
| Transformers, e come condividerlo con la community (assieme al relativo codice) così che tutte le persone possano usarlo, anche | |
| se non presente nella libreria 🤗 Transformers.</p> <p data-svelte-h="svelte-gtiyni">Illustriamo tutto questo su un modello ResNet, avvolgendo la classe ResNet della | |
| <a href="https://github.com/rwightman/pytorch-image-models" rel="nofollow">libreria timm</a> in un <code>PreTrainedModel</code>.</p> <h2 class="relative group"><a id="scrivere-una-configurazione-personalizzata" 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="#scrivere-una-configurazione-personalizzata"><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>Scrivere una configurazione personalizzata</span></h2> <p data-svelte-h="svelte-1fbquu0">Prima di iniziare a lavorare al modello, scriviamone la configurazione. La configurazione di un modello è un oggetto | |
| che contiene tutte le informazioni necessarie per la build del modello. Come vedremo nella prossima sezione, il | |
| modello può soltanto essere inizializzato tramite <code>config</code>, per cui dovremo rendere tale oggetto più completo possibile.</p> <p data-svelte-h="svelte-vu54em">Nel nostro esempio, prenderemo un paio di argomenti della classe ResNet che potremmo voler modificare. | |
| Configurazioni differenti ci daranno quindi i differenti possibili tipi di ResNet. Salveremo poi questi argomenti, | |
| dopo averne controllato la validità.</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> PretrainedConfig | |
| <span class="hljs-keyword">from</span> typing <span class="hljs-keyword">import</span> <span class="hljs-type">List</span> | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">ResnetConfig</span>(<span class="hljs-title class_ inherited__">PretrainedConfig</span>): | |
| model_type = <span class="hljs-string">"resnet"</span> | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params"> | |
| self, | |
| block_type=<span class="hljs-string">"bottleneck"</span>, | |
| layers: <span class="hljs-type">List</span>[<span class="hljs-built_in">int</span>] = [<span class="hljs-number">3</span>, <span class="hljs-number">4</span>, <span class="hljs-number">6</span>, <span class="hljs-number">3</span>], | |
| num_classes: <span class="hljs-built_in">int</span> = <span class="hljs-number">1000</span>, | |
| input_channels: <span class="hljs-built_in">int</span> = <span class="hljs-number">3</span>, | |
| cardinality: <span class="hljs-built_in">int</span> = <span class="hljs-number">1</span>, | |
| base_width: <span class="hljs-built_in">int</span> = <span class="hljs-number">64</span>, | |
| stem_width: <span class="hljs-built_in">int</span> = <span class="hljs-number">64</span>, | |
| stem_type: <span class="hljs-built_in">str</span> = <span class="hljs-string">""</span>, | |
| avg_down: <span class="hljs-built_in">bool</span> = <span class="hljs-literal">False</span>, | |
| **kwargs, | |
| </span>): | |
| <span class="hljs-keyword">if</span> block_type <span class="hljs-keyword">not</span> <span class="hljs-keyword">in</span> [<span class="hljs-string">"basic"</span>, <span class="hljs-string">"bottleneck"</span>]: | |
| <span class="hljs-keyword">raise</span> ValueError(<span class="hljs-string">f"`block_type` must be 'basic' or bottleneck', got <span class="hljs-subst">{block_type}</span>."</span>) | |
| <span class="hljs-keyword">if</span> stem_type <span class="hljs-keyword">not</span> <span class="hljs-keyword">in</span> [<span class="hljs-string">""</span>, <span class="hljs-string">"deep"</span>, <span class="hljs-string">"deep-tiered"</span>]: | |
| <span class="hljs-keyword">raise</span> ValueError(<span class="hljs-string">f"`stem_type` must be '', 'deep' or 'deep-tiered', got <span class="hljs-subst">{stem_type}</span>."</span>) | |
| self.block_type = block_type | |
| self.layers = layers | |
| self.num_classes = num_classes | |
| self.input_channels = input_channels | |
| self.cardinality = cardinality | |
| self.base_width = base_width | |
| self.stem_width = stem_width | |
| self.stem_type = stem_type | |
| self.avg_down = avg_down | |
| <span class="hljs-built_in">super</span>().__init__(**kwargs)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-pmvkz7">Le tre cose più importanti da ricordare quando scrivi le tue configurazioni sono le seguenti:</p> <ul data-svelte-h="svelte-b1zbf8"><li>Devi ereditare da <code>Pretrainedconfig</code>,</li> <li>Il metodo <code>__init__</code> del tuo <code>Pretrainedconfig</code> deve accettare i kwargs,</li> <li>I <code>kwargs</code> devono essere passati alla superclass <code>__init__</code></li></ul> <p data-svelte-h="svelte-z50i7c">L’eredità è importante per assicurarsi di ottenere tutte le funzionalità della libreria 🤗 transformers, | |
| mentre gli altri due vincoli derivano dal fatto che un <code>Pretrainedconfig</code> ha più campi di quelli che stai settando. | |
| Quando ricarichi una config da un metodo <code>from_pretrained</code>, questi campi devono essere accettati dalla tua config e | |
| poi inviati alla superclasse.</p> <p data-svelte-h="svelte-ya7kyo">Definire un <code>model_type</code> per la tua configurazione (qua <code>model_type = “resnet”</code>) non è obbligatorio, a meno che tu | |
| non voglia registrare il modello con le classi Auto (vedi l’ultima sezione).</p> <p data-svelte-h="svelte-s5azty">Una volta completato, puoi facilmente creare e salvare la tua configurazione come faresti con ogni altra configurazione | |
| di modelli della libreria. Ecco come possiamo creare la config di un resnet50d e salvarlo:</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 -->resnet50d_config = ResnetConfig(block_type=<span class="hljs-string">"bottleneck"</span>, stem_width=<span class="hljs-number">32</span>, stem_type=<span class="hljs-string">"deep"</span>, avg_down=<span class="hljs-literal">True</span>) | |
| resnet50d_config.save_pretrained(<span class="hljs-string">"custom-resnet"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-i6yebl">Questo salverà un file chiamato <code>config.json</code> all’interno della cartella <code>custom-resnet</code>. Potrai poi ricaricare la tua | |
| config con 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 -->resnet50d_config = ResnetConfig.from_pretrained(<span class="hljs-string">"custom-resnet"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-qiupgh">Puoi anche usare qualunque altro metodo della classe <code>PretrainedConfig</code>, come <code>push_to_hub()</code> | |
| per caricare direttamente la tua configurazione nell’hub.</p> <h2 class="relative group"><a id="scrivere-un-modello-personalizzato" 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="#scrivere-un-modello-personalizzato"><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>Scrivere un modello personalizzato</span></h2> <p data-svelte-h="svelte-k6id05">Ora che abbiamo la nostra configurazione ResNet, possiamo continuare a scrivere il modello. In realtà, ne scriveremo | |
| due: uno che estrae le features nascoste da una batch di immagini (come <code>BertModel</code>) e uno che è utilizzabile per | |
| la classificazione di immagini (come <code>BertModelForSequenceClassification</code>).</p> <p data-svelte-h="svelte-7jzmhs">Come abbiamo menzionato in precedenza, scriveremo soltanto un wrapper del modello, per mantenerlo semplice ai fini di | |
| questo esempio. L’unica cosa che dobbiamo fare prima di scrivere questa classe è una mappatura fra i tipi di blocco e | |
| le vere classi dei blocchi. Successivamente il modello è definito tramite la configurazione, passando tutto quanto alla | |
| classe <code>ResNet</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-keyword">from</span> transformers <span class="hljs-keyword">import</span> PreTrainedModel | |
| <span class="hljs-keyword">from</span> timm.models.resnet <span class="hljs-keyword">import</span> BasicBlock, Bottleneck, ResNet | |
| <span class="hljs-keyword">from</span> .configuration_resnet <span class="hljs-keyword">import</span> ResnetConfig | |
| BLOCK_MAPPING = {<span class="hljs-string">"basic"</span>: BasicBlock, <span class="hljs-string">"bottleneck"</span>: Bottleneck} | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">ResnetModel</span>(<span class="hljs-title class_ inherited__">PreTrainedModel</span>): | |
| config_class = ResnetConfig | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params">self, config</span>): | |
| <span class="hljs-built_in">super</span>().__init__(config) | |
| block_layer = BLOCK_MAPPING[config.block_type] | |
| self.model = ResNet( | |
| block_layer, | |
| config.layers, | |
| num_classes=config.num_classes, | |
| in_chans=config.input_channels, | |
| cardinality=config.cardinality, | |
| base_width=config.base_width, | |
| stem_width=config.stem_width, | |
| stem_type=config.stem_type, | |
| avg_down=config.avg_down, | |
| ) | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">forward</span>(<span class="hljs-params">self, tensor</span>): | |
| <span class="hljs-keyword">return</span> self.model.forward_features(tensor)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1n9eqy1">Per il modello che classificherà le immagini, cambiamo soltanto il metodo forward:</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">import</span> torch | |
| <span class="hljs-keyword">class</span> <span class="hljs-title class_">ResnetModelForImageClassification</span>(<span class="hljs-title class_ inherited__">PreTrainedModel</span>): | |
| config_class = ResnetConfig | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">__init__</span>(<span class="hljs-params">self, config</span>): | |
| <span class="hljs-built_in">super</span>().__init__(config) | |
| block_layer = BLOCK_MAPPING[config.block_type] | |
| self.model = ResNet( | |
| block_layer, | |
| config.layers, | |
| num_classes=config.num_classes, | |
| in_chans=config.input_channels, | |
| cardinality=config.cardinality, | |
| base_width=config.base_width, | |
| stem_width=config.stem_width, | |
| stem_type=config.stem_type, | |
| avg_down=config.avg_down, | |
| ) | |
| <span class="hljs-keyword">def</span> <span class="hljs-title function_">forward</span>(<span class="hljs-params">self, tensor, labels=<span class="hljs-literal">None</span></span>): | |
| logits = self.model(tensor) | |
| <span class="hljs-keyword">if</span> labels <span class="hljs-keyword">is</span> <span class="hljs-keyword">not</span> <span class="hljs-literal">None</span>: | |
| loss = torch.nn.functional.cross_entropy(logits, labels) | |
| <span class="hljs-keyword">return</span> {<span class="hljs-string">"loss"</span>: loss, <span class="hljs-string">"logits"</span>: logits} | |
| <span class="hljs-keyword">return</span> {<span class="hljs-string">"logits"</span>: logits}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1rhha2m">Nota come, in entrambi i casi, ereditiamo da <code>PreTrainedModel</code> e chiamiamo l’inizializzazione della superclasse | |
| con il metodo <code>config</code> (un po’ come quando scrivi un normale <code>torch.nn.Module</code>). La riga che imposta la <code>config_class</code> | |
| non è obbligatoria, a meno che tu non voglia registrare il modello con le classi Auto (vedi l’ultima sezione).</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-1j1jq3y">Se il tuo modello è molto simile a un modello all’interno della libreria, puoi ri-usare la stessa configurazione di quel modello.</p></div> <p data-svelte-h="svelte-1xd6e24">Puoi fare in modo che il tuo modello restituisca in output qualunque cosa tu voglia, ma far restituire un dizionario | |
| come abbiamo fatto per <code>ResnetModelForImageClassification</code>, con la funzione di perdita inclusa quando vengono passate le labels, | |
| renderà il tuo modello direttamente utilizzabile all’interno della classe <code>Trainer</code>. Utilizzare altri formati di output va bene | |
| se hai in progetto di utilizzare un tuo loop di allenamento, o se utilizzerai un’altra libreria per l’addestramento.</p> <p data-svelte-h="svelte-1eigcf8">Ora che abbiamo la classe del nostro modello, creiamone uno:</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 -->resnet50d = ResnetModelForImageClassification(resnet50d_config)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-11w45e">Ribadiamo, puoi usare qualunque metodo dei <code>PreTrainedModel</code>, come <code>save_pretrained()</code> o | |
| <code>push_to_hub()</code>. Utilizzeremo quest’ultimo nella prossima sezione, e vedremo come caricare i pesi del | |
| modello assieme al codice del modello stesso. Ma prima, carichiamo alcuni pesi pre-allenati all’interno del nostro modello.</p> <p data-svelte-h="svelte-aatwzz">Nel tuo caso specifico, probabilmente allenerai il tuo modello sui tuoi dati. Per velocizzare in questo tutorial, | |
| utilizzeremo la versione pre-allenata del resnet50d. Dato che il nostro modello è soltanto un wrapper attorno a quel modello, | |
| sarà facile trasferirne i 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-keyword">import</span> timm | |
| pretrained_model = timm.create_model(<span class="hljs-string">"resnet50d"</span>, pretrained=<span class="hljs-literal">True</span>) | |
| resnet50d.model.load_state_dict(pretrained_model.state_dict())<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-31zx9x">Vediamo adesso come assicurarci che quando facciamo <code>save_pretrained()</code> o <code>push_to_hub()</code>, | |
| il codice del modello venga salvato.</p> <h2 class="relative group"><a id="inviare-il-codice-allhub" 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="#inviare-il-codice-allhub"><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>Inviare il codice all’Hub</span></h2> <div class="course-tip course-tip-orange bg-gradient-to-br dark:bg-gradient-to-r before:border-orange-500 dark:before:border-orange-800 from-orange-50 dark:from-gray-900 to-white dark:to-gray-950 border border-orange-50 text-orange-700 dark:text-gray-400"><p data-svelte-h="svelte-1dfgrpm">Questa API è sperimentale e potrebbe avere alcuni cambiamenti nei prossimi rilasci.</p></div> <p data-svelte-h="svelte-1h1v7pd">Innanzitutto, assicurati che il tuo modello sia completamente definito in un file <code>.py</code>. Può sfruttare import relativi | |
| ad altri file, purchè questi siano nella stessa directory (non supportiamo ancora sotto-moduli per questa funzionalità). | |
| Per questo esempio, definiremo un file <code>modeling_resnet.py</code> e un file <code>configuration_resnet.py</code> in una cartella dell’attuale | |
| working directory chiamata <code>resnet_model</code>. Il file configuration contiene il codice per <code>ResnetConfig</code> e il file modeling | |
| contiene il codice di <code>ResnetModel</code> e <code>ResnetModelForImageClassification</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 -->. | |
| └── resnet_model | |
| ├── __init__.<span class="hljs-keyword">py</span> | |
| ├── configuration_resnet.<span class="hljs-keyword">py</span> | |
| └── modeling_resnet.<span class="hljs-keyword">py</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1jiyxvm">Il file <code>__init__.py</code> può essere vuoto, serve solo perchè Python capisca che <code>resnet_model</code> può essere utilizzato come un modulo.</p> <div class="course-tip course-tip-orange bg-gradient-to-br dark:bg-gradient-to-r before:border-orange-500 dark:before:border-orange-800 from-orange-50 dark:from-gray-900 to-white dark:to-gray-950 border border-orange-50 text-orange-700 dark:text-gray-400"><p data-svelte-h="svelte-ek1fx">Se stai copiando i file relativi alla modellazione della libreria, dovrai sostituire tutti gli import relativi in cima al file con import del | |
| pacchetto <code>transformers</code>.</p></div> <p data-svelte-h="svelte-lts8a1">Nota che puoi ri-utilizzare (o usare come sottoclassi) un modello/configurazione esistente.</p> <p data-svelte-h="svelte-7j1ida">Per condividere il tuo modello con la community, segui questi passi: prima importa il modello ResNet e la sua configurazione | |
| dai nuovi file creati:</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> resnet_model.configuration_resnet <span class="hljs-keyword">import</span> ResnetConfig | |
| <span class="hljs-keyword">from</span> resnet_model.modeling_resnet <span class="hljs-keyword">import</span> ResnetModel, ResnetModelForImageClassification<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-10v4ms0">Dopodichè dovrai dire alla libreria che vuoi copiare i file con il codice di quegli oggetti quando utilizzi il metodo | |
| <code>save_pretrained</code> e registrarli in modo corretto con una Auto classe (specialmente per i modelli). Utilizza semplicemente:</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 -->ResnetConfig.register_for_auto_class() | |
| ResnetModel.register_for_auto_class(<span class="hljs-string">"AutoModel"</span>) | |
| ResnetModelForImageClassification.register_for_auto_class(<span class="hljs-string">"AutoModelForImageClassification"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-16h756x">Nota che non c’è bisogno di specificare una Auto classe per la configurazione (c’è solo una Auto classe per le configurazioni, | |
| <code>AutoConfig</code>, ma è diversa per i modelli). Il tuo modello personalizato potrebbe essere utilizzato per diverse tasks, | |
| per cui devi specificare quale delle classi Auto è quella corretta per il tuo modello.</p> <p data-svelte-h="svelte-1p0sixi">Successivamente, creiamo i modelli e la config come abbiamo fatto in precedenza:</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 -->resnet50d_config = ResnetConfig(block_type=<span class="hljs-string">"bottleneck"</span>, stem_width=<span class="hljs-number">32</span>, stem_type=<span class="hljs-string">"deep"</span>, avg_down=<span class="hljs-literal">True</span>) | |
| resnet50d = ResnetModelForImageClassification(resnet50d_config) | |
| pretrained_model = timm.create_model(<span class="hljs-string">"resnet50d"</span>, pretrained=<span class="hljs-literal">True</span>) | |
| resnet50d.model.load_state_dict(pretrained_model.state_dict())<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-4z8yqk">Adesso, per inviare il modello all’Hub, assicurati di aver effettuato l’accesso. Lancia dal tuo terminale:</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 -->huggingface-cli login<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1t3y58i">O da un notebook:</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> huggingface_hub <span class="hljs-keyword">import</span> notebook_login | |
| notebook_login()<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1scrh7h">Potrai poi inviare il tutto sul tuo profilo (o di un’organizzazione di cui fai parte) in questo modo:</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 -->resnet50d.push_to_hub(<span class="hljs-string">"custom-resnet50d"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-amfesh">Oltre ai pesi del modello e alla configurazione in formato json, questo ha anche copiato i file <code>.py</code> modeling e | |
| configuration all’interno della cartella <code>custom-resnet50d</code> e ha caricato i risultati sull’Hub. Puoi controllare | |
| i risultati in questa <a href="https://huggingface.co/sgugger/custom-resnet50d" rel="nofollow">model repo</a>.</p> <p data-svelte-h="svelte-51ocuc">Puoi controllare il tutorial di condivisione <a href="model_sharing">tutorial di condivisione</a> per più informazioni sul | |
| metodo con cui inviare all’Hub.</p> <h2 class="relative group"><a id="usare-un-modello-con-codice-personalizzato" 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="#usare-un-modello-con-codice-personalizzato"><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>Usare un modello con codice personalizzato</span></h2> <p data-svelte-h="svelte-1xalseo">Puoi usare ogni configurazione, modello o tokenizer con file di codice personalizzati nella sua repository | |
| con le classi Auto e il metodo <code>from_pretrained</code>. Tutti i files e il codice caricati sull’Hub sono scansionati da malware | |
| (fai riferimento alla documentazione <a href="https://huggingface.co/docs/hub/security#malware-scanning" rel="nofollow">Hub security</a> per più informazioni), | |
| ma dovresti comunque assicurarti dell’affidabilità del codice e dell’autore per evitare di eseguire codice dannoso sulla tua macchina. | |
| Imposta <code>trust_remote_code=True</code> per usare un modello con codice personalizzato:</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> AutoModelForImageClassification | |
| model = AutoModelForImageClassification.from_pretrained(<span class="hljs-string">"sgugger/custom-resnet50d"</span>, trust_remote_code=<span class="hljs-literal">True</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-lcodn">Inoltre, raccomandiamo fortemente di passare un hash del commit come <code>revision</code> per assicurarti che le autrici o gli autori del modello | |
| non abbiano modificato il codice con alcune nuove righe dannose (a meno che non ti fidi completamente della fonte):</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 -->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 | |
| )<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-2poiu5">Nota che quando cerchi la storia dei commit della repo del modello sull’Hub, c’è un bottone con cui facilmente copiare il | |
| commit hash di ciascun commit.</p> <h2 class="relative group"><a id="registrare-un-modello-con-codice-personalizzato-nelle-classi-auto" 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="#registrare-un-modello-con-codice-personalizzato-nelle-classi-auto"><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>Registrare un modello con codice personalizzato nelle classi Auto</span></h2> <p data-svelte-h="svelte-1ucsuh2">Se stai scrivendo una libreria che estende 🤗 Transformers, potresti voler estendere le classi Auto per includere il tuo modello. | |
| Questo è diverso dall’inviare codice nell’Hub: gli utenti dovranno importare la tua libreria per ottenere il modello personalizzato | |
| (anzichè scaricare automaticamente il modello dall’Hub).</p> <p data-svelte-h="svelte-14j7ntv">Finchè il tuo file di configurazione ha un attributo <code>model_type</code> diverso dai model types esistenti, e finchè le tue | |
| classi modello hanno i corretti attributi <code>config_class</code>, potrai semplicemente aggiungerli alle classi Auto come segue:</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> AutoConfig, AutoModel, AutoModelForImageClassification | |
| AutoConfig.register(<span class="hljs-string">"resnet"</span>, ResnetConfig) | |
| AutoModel.register(ResnetConfig, ResnetModel) | |
| AutoModelForImageClassification.register(ResnetConfig, ResnetModelForImageClassification)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1a07nu7">Nota che il primo argomento utilizzato quando registri la configurazione di un modello personalizzato con <code>AutoConfig</code> | |
| deve corrispondere al <code>model_type</code> della tua configurazione personalizzata, ed il primo argomento utilizzato quando | |
| registri i tuoi modelli personalizzati in una qualunque classe Auto del modello deve corrispondere alla <code>config_class</code> | |
| di quei modelli.</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/custom_models.md" target="_blank"><span data-svelte-h="svelte-1kd6by1"><</span> <span data-svelte-h="svelte-x0xyl0">></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, 10], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 55.3 kB
- Xet hash:
- 6452b1b863eef52e3099103f4586d073937cd1098d886ef2f31a17536ab16b49
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.