trl/main/en/extras.html

<meta charset="utf-8" /><meta http-equiv="content-security-policy" content=""><meta name="hf:doc:metadata" content="{&quot;local&quot;:&quot;extras-alternative-ways-to-get-better-model-output-without-rl-based-finetuning&quot;,&quot;sections&quot;:[{&quot;local&quot;:&quot;usage&quot;,&quot;title&quot;:&quot;Usage&quot;}],&quot;title&quot;:&quot;Extras: Alternative ways to get better model output without RL based fine-tuning &quot;}" data-svelte="svelte-1phssyn">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/assets/pages/__layout.svelte-hf-doc-builder.css">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/start-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/chunks/vendor-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/chunks/paths-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/pages/__layout.svelte-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/pages/extras.mdx-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/chunks/IconCopyLink-hf-doc-builder.js">
	<link rel="modulepreload" href="/docs/trl/main/en/_app/chunks/CodeBlock-hf-doc-builder.js"> 





<h1 class="relative group"><a id="extras-alternative-ways-to-get-better-model-output-without-rl-based-finetuning" 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="#extras-alternative-ways-to-get-better-model-output-without-rl-based-finetuning"><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>Extras: Alternative ways to get better model output without RL based fine-tuning 
	</span></h1>

<p>Within the extras module is the <code>best-of-n</code> sampler class that serves as an alternative method of generating better model output.
As to how it fares against the RL based fine-tuning, please look in the <code>examples</code> directory for a comparison example</p>
<h2 class="relative group"><a id="usage" 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="#usage"><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>Usage
	</span></h2>

<p>To get started quickly, instantiate an instance of the class with a model, a length sampler, a tokenizer and a callable that serves as a proxy reward pipeline that outputs reward scores for input queries</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><!-- HTML_TAG_START -->
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> pipeline, AutoTokenizer
<span class="hljs-keyword">from</span> trl <span class="hljs-keyword">import</span> AutoModelForCausalLMWithValueHead
<span class="hljs-keyword">from</span> trl.core <span class="hljs-keyword">import</span> LengthSampler
<span class="hljs-keyword">from</span> trl.extras <span class="hljs-keyword">import</span> BestOfNSampler

ref_model = AutoModelForCausalLMWithValueHead.from_pretrained(ref_model_name)
reward_pipe = pipeline(<span class="hljs-string">&quot;sentiment-analysis&quot;</span>, model=reward_model, device=device)
tokenizer = AutoTokenizer.from_pretrained(ref_model_name)
tokenizer.pad_token = tokenizer.eos_token


<span class="hljs-comment"># callable that takes a list of raw text and returns a list of corresponding reward scores</span>
<span class="hljs-keyword">def</span> <span class="hljs-title function_">queries_to_scores</span>(<span class="hljs-params">list_of_strings</span>):
  <span class="hljs-keyword">return</span> [output[<span class="hljs-string">&quot;score&quot;</span>] <span class="hljs-keyword">for</span> output <span class="hljs-keyword">in</span> reward_pipe(list_of_strings)]

best_of_n = BestOfNSampler(model, tokenizer, queries_to_scores, length_sampler=output_length_sampler)

<!-- HTML_TAG_END --></pre></div>
<p>And assuming you have a list/tensor of tokenized queries, you can generate better output by calling the <code>generate</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><!-- HTML_TAG_START -->
best_of_n.generate(query_tensors, device=device, **gen_kwargs)
<!-- HTML_TAG_END --></pre></div>
<p>The default sample size is 4, but you can change it at the time of instance initialization like so</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><!-- HTML_TAG_START -->
best_of_n = BestOfNSampler(model, tokenizer, queries_to_scores, length_sampler=output_length_sampler, sample_size=<span class="hljs-number">8</span>)
<!-- HTML_TAG_END --></pre></div>
<p>The default output is the result of taking the top scored output for each query, but you can change it to top 2 and so on by passing the <code>n_candidates</code> argument at the time of instance initialization</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><!-- HTML_TAG_START -->
best_of_n = BestOfNSampler(model, tokenizer, queries_to_scores, length_sampler=output_length_sampler, n_candidates=<span class="hljs-number">2</span>)
<!-- HTML_TAG_END --></pre></div>
<p>There is the option of setting the generation settings (like <code>temperature</code>, <code>pad_token_id</code>) at the time of instance creation as opposed to when calling the <code>generate</code> method.
This is done by passing a <code>GenerationConfig</code> from the <code>transformers</code> library at the time of initialization</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><!-- HTML_TAG_START -->
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> GenerationConfig

generation_config = GenerationConfig(min_length= -<span class="hljs-number">1</span>, top_k=<span class="hljs-number">0.0</span>, top_p= <span class="hljs-number">1.0</span>, do_sample= <span class="hljs-literal">True</span>, pad_token_id=tokenizer.eos_token_id)

best_of_n = BestOfNSampler(model, tokenizer, queries_to_scores, length_sampler=output_length_sampler, generation_config=generation_config)

best_of_n.generate(query_tensors, device=device)
<!-- HTML_TAG_END --></pre></div>
<p>Furthermore, at the time of initialization you can set the seed to control repeatability of the generation process and the number of samples to generate for each query</p>


		<script type="module" data-hydrate="wejesa">
		import { start } from "/docs/trl/main/en/_app/start-hf-doc-builder.js";
		start({
			target: document.querySelector('[data-hydrate="wejesa"]').parentNode,
			paths: {"base":"/docs/trl/main/en","assets":"/docs/trl/main/en"},
			session: {},
			route: false,
			spa: false,
			trailing_slash: "never",
			hydrate: {
				status: 200,
				error: null,
				nodes: [
					import("/docs/trl/main/en/_app/pages/__layout.svelte-hf-doc-builder.js"),
						import("/docs/trl/main/en/_app/pages/extras.mdx-hf-doc-builder.js")
				],
				params: {}
			}
		});
	</script>