transformers/pr_36839/en/hpo_train.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Hyperparameter search&quot;,&quot;local&quot;:&quot;hyperparameter-search&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Distributed Data Parallel&quot;,&quot;local&quot;:&quot;distributed-data-parallel&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/transformers/pr_36839/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/entry/start.6be8d590.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/scheduler.01eeda35.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/singletons.177df05e.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/index.4862150a.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/paths.517376d1.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/entry/app.09748b4b.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/index.6dd51b66.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/nodes/0.8897c14d.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/nodes/29.3e145bab.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/Tip.de9bae2b.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/CodeBlock.864da1b0.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/EditOnGithub.7faefd25.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/HfOption.f7f04550.js">
		<link rel="modulepreload" href="/docs/transformers/pr_36839/en/_app/immutable/chunks/stores.318eade7.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Hyperparameter search&quot;,&quot;local&quot;:&quot;hyperparameter-search&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Distributed Data Parallel&quot;,&quot;local&quot;:&quot;distributed-data-parallel&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>   <h1 class="relative group"><a id="hyperparameter-search" 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="#hyperparameter-search"><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>Hyperparameter search</span></h1> <p data-svelte-h="svelte-w6ki0">Hyperparameter search discovers an optimal set of hyperparameters that produces the best model performance. <a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer">Trainer</a> supports several hyperparameter search backends - <a href="https://optuna.readthedocs.io/en/stable/index.html" rel="nofollow">Optuna</a>, <a href="https://docs.sigopt.com/" rel="nofollow">SigOpt</a>, <a href="https://docs.wandb.ai/" rel="nofollow">Weights &amp; Biases</a>, <a href="https://docs.ray.io/en/latest/tune/index.html" rel="nofollow">Ray Tune</a> - through  <a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer.hyperparameter_search">hyperparameter_search()</a> to optimize an objective or even multiple objectives.</p> <p data-svelte-h="svelte-14igez3">This guide will go over how to set up a hyperparameter search for each of the backends.</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 -->pip install optuna/sigopt/wandb/ray[tune]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-yemvin">To use <a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer.hyperparameter_search">hyperparameter_search()</a>, you need to create a <code>model_init</code> function. This function includes basic model information (arguments and configuration) because it needs to be reinitialized for each search trial in the run.</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-1nhxq68">The <code>model_init</code> function is incompatible with the <a href="./main_classes/trainer#transformers.Trainer.optimizers">optimizers</a> parameter. Subclass <a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer">Trainer</a> and override the <a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer.create_optimizer_and_scheduler">create_optimizer_and_scheduler()</a> method to create a custom optimizer and scheduler.</p></div> <p data-svelte-h="svelte-lepk6y">An example <code>model_init</code> function is shown below.</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">def</span> <span class="hljs-title function_">model_init</span>(<span class="hljs-params">trial</span>):
    <span class="hljs-keyword">return</span> AutoModelForSequenceClassification.from_pretrained(
        model_args.model_name_or_path,
        from_tf=<span class="hljs-built_in">bool</span>(<span class="hljs-string">&quot;.ckpt&quot;</span> <span class="hljs-keyword">in</span> model_args.model_name_or_path),
        config=config,
        cache_dir=model_args.cache_dir,
        revision=model_args.model_revision,
        token=<span class="hljs-literal">True</span> <span class="hljs-keyword">if</span> model_args.use_auth_token <span class="hljs-keyword">else</span> <span class="hljs-literal">None</span>,
    )<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-h4yjdy">Pass <code>model_init</code> to <a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer">Trainer</a> along with everything else you need for training. Then you can call <a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer.hyperparameter_search">hyperparameter_search()</a> to start the search.</p> <p data-svelte-h="svelte-187up64"><a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer.hyperparameter_search">hyperparameter_search()</a> accepts a <a href="./main_classes/trainer#transformers.Trainer.hyperparameter_search.direction">direction</a> parameter to specify whether to minimize, maximize, or minimize and maximize multiple objectives. You’ll also need to set the <a href="./main_classes/trainer#transformers.Trainer.hyperparameter_search.backend">backend</a> you’re using, an <a href="./main_classes/trainer#transformers.Trainer.hyperparameter_search.hp_space">object</a> containing the hyperparameters to optimize for, the <a href="./main_classes/trainer#transformers.Trainer.hyperparameter_search.n_trials">number of trials</a> to run, and a <a href="./main_classes/trainer#transformers.Trainer.hyperparameter_search.compute_objective">compute_objective</a> to return the objective values.</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-wa0ep5">If <a href="./main_classes/trainer#transformers.Trainer.hyperparameter_search.compute_objective">compute_objective</a> isn’t defined, the default <a href="./main_classes/trainer#transformers.Trainer.hyperparameter_search.compute_objective">compute_objective</a> is called which is the sum of an evaluation metric like F1.</p></div> <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> Trainer

trainer = Trainer(
    model=<span class="hljs-literal">None</span>,
    args=training_args,
    train_dataset=small_train_dataset,
    eval_dataset=small_eval_dataset,
    compute_metrics=compute_metrics,
    processing_class=tokenizer,
    model_init=model_init,
    data_collator=data_collator,
)
trainer.hyperparameter_search(...)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-f1b3ug">The following examples demonstrate how to perform a hyperparameter search for the learning rate and training batch size using the different backends.</p> <div class="flex space-x-2 items-center my-1.5 mr-8 h-7 !pl-0 -mx-3 md:mx-0"><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd border-gray-800 bg-black dark:bg-gray-700 text-white">Optuna </div><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd text-gray-500 cursor-pointer opacity-90 hover:text-gray-700 dark:hover:text-gray-200 hover:shadow-sm">Ray Tune </div><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd text-gray-500 cursor-pointer opacity-90 hover:text-gray-700 dark:hover:text-gray-200 hover:shadow-sm">SigOpt </div><div class="flex items-center border rounded-lg px-1.5 py-1 leading-none select-none text-smd text-gray-500 cursor-pointer opacity-90 hover:text-gray-700 dark:hover:text-gray-200 hover:shadow-sm">Weights &amp; Biases </div></div> <div class="language-select"><p data-svelte-h="svelte-eii0j4"><a href="https://optuna.readthedocs.io/en/stable/tutorial/10_key_features/002_configurations.html#sphx-glr-tutorial-10-key-features-002-configurations-py" rel="nofollow">Optuna</a> optimizes categories, integers, and floats.</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">def</span> <span class="hljs-title function_">optuna_hp_space</span>(<span class="hljs-params">trial</span>):
    <span class="hljs-keyword">return</span> {
        <span class="hljs-string">&quot;learning_rate&quot;</span>: trial.suggest_float(<span class="hljs-string">&quot;learning_rate&quot;</span>, <span class="hljs-number">1e-6</span>, <span class="hljs-number">1e-4</span>, log=<span class="hljs-literal">True</span>),
        <span class="hljs-string">&quot;per_device_train_batch_size&quot;</span>: trial.suggest_categorical(<span class="hljs-string">&quot;per_device_train_batch_size&quot;</span>, [<span class="hljs-number">16</span>, <span class="hljs-number">32</span>, <span class="hljs-number">64</span>, <span class="hljs-number">128</span>]),
    }

best_trials = trainer.hyperparameter_search(
    direction=[<span class="hljs-string">&quot;minimize&quot;</span>, <span class="hljs-string">&quot;maximize&quot;</span>],
    backend=<span class="hljs-string">&quot;optuna&quot;</span>,
    hp_space=optuna_hp_space,
    n_trials=<span class="hljs-number">20</span>,
    compute_objective=compute_objective,
)<!-- HTML_TAG_END --></pre></div>   </div>  <h2 class="relative group"><a id="distributed-data-parallel" 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="#distributed-data-parallel"><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>Distributed Data Parallel</span></h2> <p data-svelte-h="svelte-1xbn3j1"><a href="/docs/transformers/pr_36839/en/main_classes/trainer#transformers.Trainer">Trainer</a> only supports hyperparameter search for distributed data parallel (DDP) on the Optuna and SigOpt backends. Only the rank-zero process is used to generate the search trial, and the resulting parameters are passed along to the other ranks.</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/en/hpo_train.md" target="_blank"><span data-svelte-h="svelte-1kd6by1">&lt;</span> <span data-svelte-h="svelte-x0xyl0">&gt;</span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a>  <p></p> 
			
			<script>
				{
					__sveltekit_1bm5psi = {
						assets: "/docs/transformers/pr_36839/en",
						base: "/docs/transformers/pr_36839/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/transformers/pr_36839/en/_app/immutable/entry/start.6be8d590.js"),
						import("/docs/transformers/pr_36839/en/_app/immutable/entry/app.09748b4b.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 29],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>