Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Quicktour","local":"quicktour","sections":[{"title":"PEFT configuration and model","local":"peft-configuration-and-model","sections":[{"title":"Save model","local":"save-model","sections":[],"depth":3}],"depth":2},{"title":"Inference","local":"inference","sections":[],"depth":2},{"title":"Multiple adapters","local":"multiple-adapters","sections":[],"depth":2},{"title":"Next steps","local":"next-steps","sections":[],"depth":2}],"depth":1}"> | |
| <link href="/docs/peft/pr_3270/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/entry/start.f28557bc.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/scheduler.78382b47.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/singletons.2ac5b46a.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/index.fadd215c.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/paths.992e82e0.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/entry/app.2273f60f.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/preload-helper.251eab2a.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/index.6dd35eb6.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/nodes/0.aaf7e185.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/nodes/74.99a4177e.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/CopyLLMTxtMenu.ddc04658.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.3b3e5fc5.js"> | |
| <link rel="modulepreload" href="/docs/peft/pr_3270/en/_app/immutable/chunks/CodeBlock.a62ab26b.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Quicktour","local":"quicktour","sections":[{"title":"PEFT configuration and model","local":"peft-configuration-and-model","sections":[{"title":"Save model","local":"save-model","sections":[],"depth":3}],"depth":2},{"title":"Inference","local":"inference","sections":[],"depth":2},{"title":"Multiple adapters","local":"multiple-adapters","sections":[],"depth":2},{"title":"Next steps","local":"next-steps","sections":[],"depth":2}],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 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> </div> <h1 class="relative group"><a id="quicktour" 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="#quicktour"><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>Quicktour</span></h1> <p data-svelte-h="svelte-hgsptz">PEFT offers parameter-efficient methods for finetuning large pretrained models. The traditional paradigm is to finetune all of a model’s parameters for each downstream task, but this is becoming exceedingly costly and impractical because of the enormous number of parameters in models today. Instead, it is more efficient to train a smaller number of prompt parameters or use a reparametrization method like low-rank adaptation (LoRA) to reduce the number of trainable parameters.</p> <div class="flex justify-center" data-svelte-h="svelte-tlm57b"><div class="flex flex-col basis-1/4 pt-5"><i>PEFT can be thought of as a framework for adding trainable parameters to arbitrary places in existing models ("base models"). Specific PEFT methods arrange the trainable parameters in certain ways or modify the training process to achieve fine-tuning performance comparable to training all parameters of the base model.</i></div> <div class="flex flex-col basis-3/4 pl-10 pr-10"><img style="border: 0;box-shadow: none;border-radius: 0;" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/peft/adapter_installation.png" width="100%"></div></div> <p data-svelte-h="svelte-1t6b3ei">This quicktour will show you PEFT’s main features and how you can train or run inference on large models that would typically be inaccessible on consumer devices.</p> <h2 class="relative group"><a id="peft-configuration-and-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="#peft-configuration-and-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>PEFT configuration and model</span></h2> <p data-svelte-h="svelte-1kfg66g">For any PEFT method, you’ll need to create a configuration which contains all the parameters that specify how the PEFT method should be applied, most importantly which layers of the existing model to target with trainable parameters. Once the configuration is setup, pass it to the <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.get_peft_model">get_peft_model()</a> function along with the base model to create a trainable <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.PeftModel">PeftModel</a>.</p> <p data-svelte-h="svelte-f5dj8h">Let’s use <a href="./package_reference/lora">LoRA</a> as an example but only discuss common parameters - you might want to use one of the <a href="./methods/overview">many other PEFT methods</a>. | |
| The configuration usually entails this:</p> <ul data-svelte-h="svelte-1dvwwdp"><li><code>target_modules</code>: which modules of the base model to adapt</li> <li><code>task_type</code> (default: <code>None</code>, see <a href="package_reference/peft_types#peft.TaskType">available <code>TaskType</code>s</a>): the nature of the trained task; if provided may help to automatically save relevant layers alongside the adapter weights or warn you about incompatibilities</li> <li><code>inference_mode</code> (default: <code>False</code>): whether you’re using the model for inference or not</li></ul> <p data-svelte-h="svelte-2pkb8w">Depending on the PEFT method you choose you will add specific parameters that, for example, determine the size of the update matrices. | |
| Here’s an example of a config you may encounter in the wild:</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="language-python "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> peft <span class="hljs-keyword">import</span> LoraConfig, TaskType | |
| peft_config = LoraConfig(target_modules=[<span class="hljs-string">"q_proj"</span>], task_type=TaskType.CAUSAL_LM, inference_mode=<span class="hljs-literal">False</span>, r=<span class="hljs-number">8</span>, lora_alpha=<span class="hljs-number">32</span>, lora_dropout=<span class="hljs-number">0.1</span>)<!-- HTML_TAG_END --></pre></div> <blockquote class="tip" data-svelte-h="svelte-vy7eoy"><p>See the <a href="guides/peft_model_config">configuration guide</a> for more details on how the PEFT configuration works under the hood.</p></blockquote> <p data-svelte-h="svelte-jba0j7">Once the <a href="/docs/peft/pr_3270/en/package_reference/lora#peft.LoraConfig">LoraConfig</a> is set up, create a <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.PeftModel">PeftModel</a> with the <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.get_peft_model">get_peft_model()</a> function. It takes a base model - which you can (but don’t have to) load from the Transformers library - and the <a href="/docs/peft/pr_3270/en/package_reference/lora#peft.LoraConfig">LoraConfig</a> containing the parameters for how to configure a model for training with LoRA.</p> <p data-svelte-h="svelte-toa5ud">Load the base model you want to finetune.</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="language-python "><!-- HTML_TAG_START --><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-3.2-1B"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-padu1">Now wrap the base model and <code>peft_config</code> with the <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.get_peft_model">get_peft_model()</a> function to create a <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.PeftModel">PeftModel</a>.</p> <div class="flex justify-center" data-svelte-h="svelte-ezvah5"><div class="flex flex-col basis-2/4"><p>Wrapping means that PEFT replaces the targeted layers (here: all <code>q_proj</code> layers) with the adapter-specific layer for the target layer's type. Since we're dealing with linear layers, it will be, in this case, a <code>lora.Linear</code> layer. <b>Note</b> that these changes are done in-place to save memory, so your base model is now modified.</p> <p>Note that we've only specified <code>q_proj</code> but in actuality we are targeting all <code>model.layers[:].self_attn.q_proj</code> layers. This is because PEFT searches for matching suffixes by default. Pass a string with a regular expression if you want to target more complex layer patterns.</p></div> <div class="flex flex-col basis-2/4 pl-10 pr-10"><img style="border: 0;box-shadow: none;border-radius: 0;" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/peft/adapter_attention_targeting.png" width="600px"></div></div> <div class="flex justify-center" data-svelte-h="svelte-1635jw6"><div class="flex flex-col basis-2/4 pr-10"><img style="border: 0;box-shadow: none;border-radius: 0;" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/peft/adapter_layer_wrapping.png" width="1000px"></div> <div class="flex flex-col basis-2/4"><p>The base model's layer will be wrapped, retained and not trained while new, trainable weights are added and are combined. How these new weights are structured and combined with the weights of the base model is a good portion of what sets the different PEFT methods apart.</p></div></div> <p data-svelte-h="svelte-1vd4vp6">To get a sense of the number of trainable parameters in your model, use the <code>print_trainable_parameters</code> method.</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="language-python "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> peft <span class="hljs-keyword">import</span> get_peft_model | |
| peft_model = get_peft_model(model, peft_config) | |
| peft_model.print_trainable_parameters() | |
| <span class="hljs-string">"output: trainable params: 524,288 || all params: 1,236,338,688 || trainable%: 0.0424"</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-n91eid">Out of <a href="https://huggingface.co/meta-llama/Llama-3.2-1B" rel="nofollow">meta-llama/Llama-3.2-1B’s</a> 1B parameters, you’re only training 0.04% of them!</p> <p data-svelte-h="svelte-14m052t">That is it 🎉! Now you can train the model with the Transformers <a href="https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.Trainer" rel="nofollow">Trainer</a>, Accelerate, or any custom PyTorch training loop.</p> <p data-svelte-h="svelte-1kytgnd">For example, to train with the <a href="https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.Trainer" rel="nofollow">Trainer</a> class, setup a <a href="https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.TrainingArguments" rel="nofollow">TrainingArguments</a> class with some training hyperparameters.</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="language-py "><!-- HTML_TAG_START -->training_args = TrainingArguments( | |
| output_dir=<span class="hljs-string">"your-name/meta-llama/my-llama3.2-adapter"</span>, | |
| learning_rate=<span class="hljs-number">1e-3</span>, | |
| per_device_train_batch_size=<span class="hljs-number">32</span>, | |
| per_device_eval_batch_size=<span class="hljs-number">32</span>, | |
| num_train_epochs=<span class="hljs-number">2</span>, | |
| weight_decay=<span class="hljs-number">0.01</span>, | |
| eval_strategy=<span class="hljs-string">"epoch"</span>, | |
| save_strategy=<span class="hljs-string">"epoch"</span>, | |
| load_best_model_at_end=<span class="hljs-literal">True</span>, | |
| )<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-15wg15i">Pass the model, training arguments, dataset, tokenizer, and any other necessary component to the <a href="https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.Trainer" rel="nofollow">Trainer</a>, and call <a href="https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.Trainer.train" rel="nofollow">train</a> to start training.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class="language-py "><!-- HTML_TAG_START -->trainer = Trainer( | |
| model=peft_model, | |
| args=training_args, | |
| train_dataset=tokenized_datasets[<span class="hljs-string">"train"</span>], | |
| eval_dataset=tokenized_datasets[<span class="hljs-string">"test"</span>], | |
| data_collator=data_collator, | |
| compute_metrics=compute_metrics, | |
| ) | |
| trainer.train()<!-- HTML_TAG_END --></pre></div> <h3 class="relative group"><a id="save-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="#save-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>Save model</span></h3> <p data-svelte-h="svelte-1qx9cqx">After your model is finished training, you can save your model to a directory using the <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.PeftModel.save_pretrained">save_pretrained()</a> function.</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="language-py "><!-- HTML_TAG_START -->peft_model.save_pretrained(<span class="hljs-string">"output_dir"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-xdfp0a">You can also save your model to the Hub (make sure you’re logged in to your Hugging Face account first) with the <a href="https://huggingface.co/docs/transformers/main/en/main_classes/model#transformers.PreTrainedModel.push_to_hub" rel="nofollow">push_to_hub</a> function.</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="language-python "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> huggingface_hub <span class="hljs-keyword">import</span> notebook_login | |
| notebook_login() | |
| peft_model.push_to_hub(<span class="hljs-string">"your-name/my-llama3.2-adapter"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-lvoeik">Both methods only save the extra PEFT weights that were trained, meaning it is super efficient to store, transfer, and load. For example, this <a href="https://huggingface.co/ybelkada/opt-350m-lora" rel="nofollow">facebook/opt-350m</a> model trained with LoRA only contains two files: <code>adapter_config.json</code> and <code>adapter_model.safetensors</code>. The <code>adapter_model.safetensors</code> file is just 6.3MB!</p> <div class="flex flex-col justify-center" data-svelte-h="svelte-1nqpszv"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/peft/PEFT-hub-screenshot.png"> <figcaption class="text-center">The adapter weights for a opt-350m model stored on the Hub are only ~6MB compared to the full size of the model weights, which can be ~700MB.</figcaption></div> <h2 class="relative group"><a id="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="#inference"><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>Inference</span></h2> <blockquote class="tip" data-svelte-h="svelte-147lts3"><p>Take a look at the <a href="package_reference/auto_class">AutoPeftModel</a> API reference for a complete list of available <code>AutoPeftModel</code> classes.</p></blockquote> <p data-svelte-h="svelte-1i312rm">Easily load any PEFT-trained model for inference with the <a href="/docs/peft/pr_3270/en/package_reference/auto_class#peft.AutoPeftModel">AutoPeftModel</a> class and the <a href="https://huggingface.co/docs/transformers/main/en/main_classes/model#transformers.PreTrainedModel.from_pretrained" rel="nofollow">from_pretrained</a> method:</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="language-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> peft <span class="hljs-keyword">import</span> AutoPeftModelForCausalLM | |
| <span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer | |
| <span class="hljs-keyword">import</span> torch | |
| peft_model = AutoPeftModelForCausalLM.from_pretrained(<span class="hljs-string">"ybelkada/opt-350m-lora"</span>) | |
| tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"facebook/opt-350m"</span>) | |
| peft_model = peft_model.to(<span class="hljs-string">"cuda"</span>) | |
| peft_model.<span class="hljs-built_in">eval</span>() | |
| inputs = tokenizer(<span class="hljs-string">"Preheat the oven to 350 degrees and place the cookie dough"</span>, return_tensors=<span class="hljs-string">"pt"</span>) | |
| outputs = peft_model.generate(input_ids=inputs[<span class="hljs-string">"input_ids"</span>].to(<span class="hljs-string">"cuda"</span>), max_new_tokens=<span class="hljs-number">50</span>) | |
| <span class="hljs-built_in">print</span>(tokenizer.batch_decode(outputs.detach().cpu().numpy(), skip_special_tokens=<span class="hljs-literal">True</span>)[<span class="hljs-number">0</span>]) | |
| <span class="hljs-string">"Preheat the oven to 350 degrees and place the cookie dough in the center of the oven. In a large bowl, combine the flour, baking powder, baking soda, salt, and cinnamon. In a separate bowl, combine the egg yolks, sugar, and vanilla."</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-z24g5m">For other tasks that aren’t explicitly supported with an <code>AutoPeftModelFor</code> class - such as automatic speech recognition - you can still use the base <a href="/docs/peft/pr_3270/en/package_reference/auto_class#peft.AutoPeftModel">AutoPeftModel</a> class to load a model for the 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 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="language-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> peft <span class="hljs-keyword">import</span> AutoPeftModel | |
| peft_model = AutoPeftModel.from_pretrained(<span class="hljs-string">"smangrul/openai-whisper-large-v2-LORA-colab"</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1d61l7f">The most general way of loading a trained PEFT adapter onto a model is to use <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.PeftModel.from_pretrained">from_pretrained()</a>:</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class="language-py "><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoPeftModelForCausalLM | |
| <span class="hljs-keyword">from</span> peft <span class="hljs-keyword">import</span> PeftModel | |
| base_model = AutoModelForCausalLM.from_pretrained(<span class="hljs-string">"meta-llama/llama3.2-1B"</span>) | |
| peft_model = PeftModel.from_pretrained(base_model, <span class="hljs-string">"my-user/my-llama-adapter"</span>) <span class="hljs-comment"># you can also pass a directory instead of a hub path</span><!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="multiple-adapters" 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="#multiple-adapters"><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>Multiple adapters</span></h2> <p data-svelte-h="svelte-1crwvlu">PEFT supports installing multiple adapters (of the same kind, in this document this would be LoRA) on top of a base model. When you call <code>get_peft_model</code> there is only one adapter named <code>"default"</code> but you can add as many additional adapters as you want by calling <code>peft_model.add_adapter(adapter_name=...)</code>.</p> <div class="flex justify-center" data-svelte-h="svelte-crbgww"><div class="flex flex-col basis-2/4"><p>This works because the wrapped layer actually has a unique set of trainable weights for each adapter name. Not every adapter is active and trainable by default. You have to explicitly enable adapters by name before they are active. This allows you to quickly swap between adapters where task-specific knowledge is needed or serve different use-cases on top of one model.</p></div> <div class="flex flex-col basis-2/4 pl-10 pr-10"><img style="border: 0;box-shadow: none;border-radius: 0;" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/peft/adapter_layer_wrapping_multi_adapter.png" width="1000px"></div></div> <p data-svelte-h="svelte-lv72tz">Just remember to call <code>peft_model.set_adapter(<adapter_name>)</code> first to enable the adapter.</p> <p data-svelte-h="svelte-1y36cnf">Quick example:</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="language-py "><!-- HTML_TAG_START -->peft_model.add_adapter(adapter_name=<span class="hljs-string">'new_adapter'</span>) | |
| peft_model.set_adapter(<span class="hljs-string">'new_adapter'</span>)<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="next-steps" 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="#next-steps"><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>Next steps</span></h2> <p data-svelte-h="svelte-zexwzv">Now that you’ve seen how to train a model with one of the PEFT methods, we encourage you to try out some of the other methods like prompt tuning. The steps are very similar to the ones shown in the quicktour:</p> <ol data-svelte-h="svelte-1r4ml46"><li>prepare a <a href="/docs/peft/pr_3270/en/package_reference/config#peft.PeftConfig">PeftConfig</a> for a PEFT method, e.g. a <a href="/docs/peft/pr_3270/en/package_reference/lora#peft.LoraConfig">LoraConfig</a> or some other config (see the <a href="methods/overview">method overview</a>)</li> <li>use the <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.get_peft_model">get_peft_model()</a> method to create a <a href="/docs/peft/pr_3270/en/package_reference/peft_model#peft.PeftModel">PeftModel</a> from the configuration and base model</li></ol> <p data-svelte-h="svelte-npklm6">Then you can train it however you like! To load a PEFT model for inference, you can use the <a href="/docs/peft/pr_3270/en/package_reference/auto_class#peft.AutoPeftModel">AutoPeftModel</a> class.</p> <p data-svelte-h="svelte-eanbvx">Feel free to also take a look at the task guides if you’re interested in training a model with another PEFT method for a specific task such as semantic segmentation, multilingual automatic speech recognition, DreamBooth, token classification, and more.</p> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/peft/blob/main/docs/source/quicktour.md" target="_blank"><svg class="mr-1" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg> <span data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a> <p></p> | |
| <script> | |
| { | |
| __sveltekit_1xnt4vv = { | |
| assets: "/docs/peft/pr_3270/en", | |
| base: "/docs/peft/pr_3270/en", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/peft/pr_3270/en/_app/immutable/entry/start.f28557bc.js"), | |
| import("/docs/peft/pr_3270/en/_app/immutable/entry/app.2273f60f.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 74], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 44.1 kB
- Xet hash:
- a649488f3b70183c160cfb0fc80cf67eb0ab58024e867da1efa3b1decafa35cc
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.