Buckets:

rtrm's picture
download
raw
39 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Perplexity of fixed-length models&quot;,&quot;local&quot;:&quot;perplexity-of-fixed-length-models&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Calculating PPL with fixed-length models&quot;,&quot;local&quot;:&quot;calculating-ppl-with-fixed-length-models&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Example: Calculating perplexity with GPT-2 in 🤗 Transformers&quot;,&quot;local&quot;:&quot;example-calculating-perplexity-with-gpt-2-in--transformers&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
<link href="/docs/transformers/pr_33913/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/entry/start.b67f883f.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/scheduler.25b97de1.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/singletons.62a184e0.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/index.e188933d.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/paths.51881b9e.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/entry/app.e436b1f2.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/index.d9030fc9.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/nodes/0.05e395f5.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/each.e59479a4.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/nodes/390.49876948.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/CodeBlock.e6cd0d95.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/DocNotebookDropdown.5ea6cb78.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/globals.7f7f1b26.js">
<link rel="modulepreload" href="/docs/transformers/pr_33913/en/_app/immutable/chunks/EditOnGithub.91d95064.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Perplexity of fixed-length models&quot;,&quot;local&quot;:&quot;perplexity-of-fixed-length-models&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Calculating PPL with fixed-length models&quot;,&quot;local&quot;:&quot;calculating-ppl-with-fixed-length-models&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Example: Calculating perplexity with GPT-2 in 🤗 Transformers&quot;,&quot;local&quot;:&quot;example-calculating-perplexity-with-gpt-2-in--transformers&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="perplexity-of-fixed-length-models" 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="#perplexity-of-fixed-length-models"><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>Perplexity of fixed-length models</span></h1> <div class="flex space-x-1 absolute z-10 right-0 top-0"> <div class="relative colab-dropdown "> <button class=" " type="button"> <img alt="Open In Colab" class="!m-0" src="https://colab.research.google.com/assets/colab-badge.svg"> </button> </div> <div class="relative colab-dropdown "> <button class=" " type="button"> <img alt="Open In Studio Lab" class="!m-0" src="https://studiolab.sagemaker.aws/studiolab.svg"> </button> </div></div> <p data-svelte-h="svelte-syn4xn">Perplexity (PPL) is one of the most common metrics for evaluating language models. Before diving in, we should note
that the metric applies specifically to classical language models (sometimes called autoregressive or causal language
models) and is not well defined for masked language models like BERT (see <a href="model_summary">summary of the models</a>).</p> <p>Perplexity is defined as the exponentiated average negative log-likelihood of a sequence. If we have a tokenized
sequence<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>X</mi><mo>=</mo><mo stretchy="false">(</mo><msub><mi>x</mi><mn>0</mn></msub><mo separator="true">,</mo><msub><mi>x</mi><mn>1</mn></msub><mo separator="true">,</mo><mo></mo><mo separator="true">,</mo><msub><mi>x</mi><mi>t</mi></msub><mo stretchy="false">)</mo></mrow><annotation encoding="application/x-tex">X = (x_0, x_1, \dots, x_t)</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.6833em;"></span><span class="mord mathnormal" style="margin-right:0.07847em;">X</span><span class="mspace" style="margin-right:0.2778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2778em;"></span></span><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3011em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight">0</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mpunct">,</span><span class="mspace" style="margin-right:0.1667em;"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3011em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mpunct">,</span><span class="mspace" style="margin-right:0.1667em;"></span><span class="minner"></span><span class="mspace" style="margin-right:0.1667em;"></span><span class="mpunct">,</span><span class="mspace" style="margin-right:0.1667em;"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2806em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">t</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mclose">)</span></span></span></span><!-- HTML_TAG_END -->, then the perplexity of<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>X</mi></mrow><annotation encoding="application/x-tex">X</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.6833em;"></span><span class="mord mathnormal" style="margin-right:0.07847em;">X</span></span></span></span><!-- HTML_TAG_END --> is,
<!-- HTML_TAG_START --><span class="katex-display"><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML" display="block"><semantics><mrow><mtext>PPL</mtext><mo stretchy="false">(</mo><mi>X</mi><mo stretchy="false">)</mo><mo>=</mo><mi>exp</mi><mo></mo><mrow><mo fence="true">{</mo><mrow><mo></mo><mfrac><mn>1</mn><mi>t</mi></mfrac><munderover><mo></mo><mi>i</mi><mi>t</mi></munderover><mi>log</mi><mo></mo><msub><mi>p</mi><mi>θ</mi></msub><mo stretchy="false">(</mo><msub><mi>x</mi><mi>i</mi></msub><mi mathvariant="normal"></mi><msub><mi>x</mi><mrow><mo>&lt;</mo><mi>i</mi></mrow></msub><mo stretchy="false">)</mo></mrow><mo fence="true">}</mo></mrow></mrow><annotation encoding="application/x-tex">\text{PPL}(X) = \exp \left\{ {-\frac{1}{t}\sum_i^t \log p_\theta (x_i|x_{&lt;i}) } \right\}</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord text"><span class="mord">PPL</span></span><span class="mopen">(</span><span class="mord mathnormal" style="margin-right:0.07847em;">X</span><span class="mclose">)</span><span class="mspace" style="margin-right:0.2778em;"></span><span class="mrel">=</span><span class="mspace" style="margin-right:0.2778em;"></span></span><span class="base"><span class="strut" style="height:3.0582em;vertical-align:-1.2777em;"></span><span class="mop">exp</span><span class="mspace" style="margin-right:0.1667em;"></span><span class="minner"><span class="mopen delimcenter" style="top:0em;"><span class="delimsizing size4">{</span></span><span class="mord"><span class="mord"></span><span class="mord"><span class="mopen nulldelimiter"></span><span class="mfrac"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:1.3214em;"><span style="top:-2.314em;"><span class="pstrut" style="height:3em;"></span><span class="mord"><span class="mord mathnormal">t</span></span></span><span style="top:-3.23em;"><span class="pstrut" style="height:3em;"></span><span class="frac-line" style="border-bottom-width:0.04em;"></span></span><span style="top:-3.677em;"><span class="pstrut" style="height:3em;"></span><span class="mord"><span class="mord">1</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.686em;"><span></span></span></span></span></span><span class="mclose nulldelimiter"></span></span><span class="mspace" style="margin-right:0.1667em;"></span><span class="mop op-limits"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:1.7806em;"><span style="top:-1.8723em;margin-left:0em;"><span class="pstrut" style="height:3.05em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">i</span></span></span><span style="top:-3.05em;"><span class="pstrut" style="height:3.05em;"></span><span><span class="mop op-symbol large-op"></span></span></span><span style="top:-4.3em;margin-left:0em;"><span class="pstrut" style="height:3.05em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">t</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:1.2777em;"><span></span></span></span></span></span><span class="mspace" style="margin-right:0.1667em;"></span><span class="mop">lo<span style="margin-right:0.01389em;">g</span></span><span class="mspace" style="margin-right:0.1667em;"></span><span class="mord"><span class="mord mathnormal">p</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3361em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight" style="margin-right:0.02778em;">θ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3117em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">i</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mord"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3117em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight"><span class="mrel mtight">&lt;</span><span class="mord mathnormal mtight">i</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.1774em;"><span></span></span></span></span></span></span><span class="mclose">)</span></span><span class="mclose delimcenter" style="top:0em;"><span class="delimsizing size4">}</span></span></span></span></span></span></span><!-- HTML_TAG_END --></p> <p>where<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>log</mi><mo></mo><msub><mi>p</mi><mi>θ</mi></msub><mo stretchy="false">(</mo><msub><mi>x</mi><mi>i</mi></msub><mi mathvariant="normal"></mi><msub><mi>x</mi><mrow><mo>&lt;</mo><mi>i</mi></mrow></msub><mo stretchy="false">)</mo></mrow><annotation encoding="application/x-tex">\log p_\theta (x_i|x_{&lt;i})</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mop">lo<span style="margin-right:0.01389em;">g</span></span><span class="mspace" style="margin-right:0.1667em;"></span><span class="mord"><span class="mord mathnormal">p</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3361em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight" style="margin-right:0.02778em;">θ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3117em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">i</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mord"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3117em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight"><span class="mrel mtight">&lt;</span><span class="mord mathnormal mtight">i</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.1774em;"><span></span></span></span></span></span></span><span class="mclose">)</span></span></span></span><!-- HTML_TAG_END --> is the log-likelihood of the ith token conditioned on the preceding tokens<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><msub><mi>x</mi><mrow><mo>&lt;</mo><mi>i</mi></mrow></msub></mrow><annotation encoding="application/x-tex">x_{&lt;i}</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.6079em;vertical-align:-0.1774em;"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3117em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight"><span class="mrel mtight">&lt;</span><span class="mord mathnormal mtight">i</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.1774em;"><span></span></span></span></span></span></span></span></span></span><!-- HTML_TAG_END --> according to our model. Intuitively, it can be thought of as an evaluation of the model’s ability to predict uniformly among the set of specified tokens in a corpus. Importantly, this means that the tokenization procedure has a direct impact on a model’s perplexity which should always be taken into consideration when comparing different models.</p> <p data-svelte-h="svelte-10yo7ds">This is also equivalent to the exponentiation of the cross-entropy between the data and model predictions. For more
intuition about perplexity and its relationship to Bits Per Character (BPC) and data compression, check out this
<a href="https://thegradient.pub/understanding-evaluation-metrics-for-language-models/" rel="nofollow">fantastic blog post on The Gradient</a>.</p> <h2 class="relative group"><a id="calculating-ppl-with-fixed-length-models" 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="#calculating-ppl-with-fixed-length-models"><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>Calculating PPL with fixed-length models</span></h2> <p data-svelte-h="svelte-nt0thp">If we weren’t limited by a model’s context size, we would evaluate the model’s perplexity by autoregressively
factorizing a sequence and conditioning on the entire preceding subsequence at each step, as shown below.</p> <img width="600" alt="Full decomposition of a sequence with unlimited context length" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/ppl_full.gif"> <p>When working with approximate models, however, we typically have a constraint on the number of tokens the model can
process. The largest version of <a href="model_doc/gpt2" data-svelte-h="svelte-1kdeo4m">GPT-2</a>, for example, has a fixed length of 1024 tokens, so we
cannot calculate<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><msub><mi>p</mi><mi>θ</mi></msub><mo stretchy="false">(</mo><msub><mi>x</mi><mi>t</mi></msub><mi mathvariant="normal"></mi><msub><mi>x</mi><mrow><mo>&lt;</mo><mi>t</mi></mrow></msub><mo stretchy="false">)</mo></mrow><annotation encoding="application/x-tex">p_\theta(x_t|x_{&lt;t})</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:1em;vertical-align:-0.25em;"></span><span class="mord"><span class="mord mathnormal">p</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.3361em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight" style="margin-right:0.02778em;">θ</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mopen">(</span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2806em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">t</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span><span class="mord"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2806em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mtight"><span class="mrel mtight">&lt;</span><span class="mord mathnormal mtight">t</span></span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.1774em;"><span></span></span></span></span></span></span><span class="mclose">)</span></span></span></span><!-- HTML_TAG_END --> directly when<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>t</mi></mrow><annotation encoding="application/x-tex">t</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.6151em;"></span><span class="mord mathnormal">t</span></span></span></span><!-- HTML_TAG_END --> is greater than 1024.</p> <p>Instead, the sequence is typically broken into subsequences equal to the model’s maximum input size. If a model’s max
input size is<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>k</mi></mrow><annotation encoding="application/x-tex">k</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.6944em;"></span><span class="mord mathnormal" style="margin-right:0.03148em;">k</span></span></span></span><!-- HTML_TAG_END -->, we then approximate the likelihood of a token<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><msub><mi>x</mi><mi>t</mi></msub></mrow><annotation encoding="application/x-tex">x_t</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.5806em;vertical-align:-0.15em;"></span><span class="mord"><span class="mord mathnormal">x</span><span class="msupsub"><span class="vlist-t vlist-t2"><span class="vlist-r"><span class="vlist" style="height:0.2806em;"><span style="top:-2.55em;margin-left:0em;margin-right:0.05em;"><span class="pstrut" style="height:2.7em;"></span><span class="sizing reset-size6 size3 mtight"><span class="mord mathnormal mtight">t</span></span></span></span><span class="vlist-s"></span></span><span class="vlist-r"><span class="vlist" style="height:0.15em;"><span></span></span></span></span></span></span></span></span></span><!-- HTML_TAG_END --> by conditioning only on the<!-- HTML_TAG_START --><span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>k</mi><mo></mo><mn>1</mn></mrow><annotation encoding="application/x-tex">k-1</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.7778em;vertical-align:-0.0833em;"></span><span class="mord mathnormal" style="margin-right:0.03148em;">k</span><span class="mspace" style="margin-right:0.2222em;"></span><span class="mbin"></span><span class="mspace" style="margin-right:0.2222em;"></span></span><span class="base"><span class="strut" style="height:0.6444em;"></span><span class="mord">1</span></span></span></span><!-- HTML_TAG_END --> tokens that precede it rather than the entire context. When evaluating the model’s perplexity of a
sequence, a tempting but suboptimal approach is to break the sequence into disjoint chunks and add up the decomposed
log-likelihoods of each segment independently.</p> <img width="600" alt="Suboptimal PPL not taking advantage of full available context" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/ppl_chunked.gif"> <p data-svelte-h="svelte-19oj8tf">This is quick to compute since the perplexity of each segment can be computed in one forward pass, but serves as a poor
approximation of the fully-factorized perplexity and will typically yield a higher (worse) PPL because the model will
have less context at most of the prediction steps.</p> <p data-svelte-h="svelte-w5z9wi">Instead, the PPL of fixed-length models should be evaluated with a sliding-window strategy. This involves repeatedly
sliding the context window so that the model has more context when making each prediction.</p> <img width="600" alt="Sliding window PPL taking advantage of all available context" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/ppl_sliding.gif"> <p data-svelte-h="svelte-zqbk5w">This is a closer approximation to the true decomposition of the sequence probability and will typically yield a more
favorable score. The downside is that it requires a separate forward pass for each token in the corpus. A good
practical compromise is to employ a strided sliding window, moving the context by larger strides rather than sliding by
1 token a time. This allows computation to proceed much faster while still giving the model a large context to make
predictions at each step.</p> <h2 class="relative group"><a id="example-calculating-perplexity-with-gpt-2-in--transformers" 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="#example-calculating-perplexity-with-gpt-2-in--transformers"><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>Example: Calculating perplexity with GPT-2 in 🤗 Transformers</span></h2> <p data-svelte-h="svelte-atwk84">Let’s demonstrate this process with GPT-2.</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> GPT2LMHeadModel, GPT2TokenizerFast
device = <span class="hljs-string">&quot;cuda&quot;</span>
model_id = <span class="hljs-string">&quot;openai-community/gpt2-large&quot;</span>
model = GPT2LMHeadModel.from_pretrained(model_id).to(device)
tokenizer = GPT2TokenizerFast.from_pretrained(model_id)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1jth08f">We’ll load in the WikiText-2 dataset and evaluate the perplexity using a few different sliding-window strategies. Since
this dataset is small and we’re just doing one forward pass over the set, we can just load and encode the entire
dataset in memory.</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> datasets <span class="hljs-keyword">import</span> load_dataset
test = load_dataset(<span class="hljs-string">&quot;wikitext&quot;</span>, <span class="hljs-string">&quot;wikitext-2-raw-v1&quot;</span>, split=<span class="hljs-string">&quot;test&quot;</span>)
encodings = tokenizer(<span class="hljs-string">&quot;\n\n&quot;</span>.join(test[<span class="hljs-string">&quot;text&quot;</span>]), return_tensors=<span class="hljs-string">&quot;pt&quot;</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1jepp38">With 🤗 Transformers, we can simply pass the <code>input_ids</code> as the <code>labels</code> to our model, and the average negative
log-likelihood for each token is returned as the loss. With our sliding window approach, however, there is overlap in
the tokens we pass to the model at each iteration. We don’t want the log-likelihood for the tokens we’re just treating
as context to be included in our loss, so we can set these targets to <code>-100</code> so that they are ignored. The following
is an example of how we could do this with a stride of <code>512</code>. This means that the model will have at least 512 tokens
for context when calculating the conditional likelihood of any one token (provided there are 512 preceding tokens
available to condition on).</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">from</span> tqdm <span class="hljs-keyword">import</span> tqdm
max_length = model.config.n_positions
stride = <span class="hljs-number">512</span>
seq_len = encodings.input_ids.size(<span class="hljs-number">1</span>)
nll_sum = <span class="hljs-number">0.0</span>
n_tokens = <span class="hljs-number">0</span>
prev_end_loc = <span class="hljs-number">0</span>
<span class="hljs-keyword">for</span> begin_loc <span class="hljs-keyword">in</span> tqdm(<span class="hljs-built_in">range</span>(<span class="hljs-number">0</span>, seq_len, stride)):
end_loc = <span class="hljs-built_in">min</span>(begin_loc + max_length, seq_len)
trg_len = end_loc - prev_end_loc <span class="hljs-comment"># may be different from stride on last loop</span>
input_ids = encodings.input_ids[:, begin_loc:end_loc].to(device)
target_ids = input_ids.clone()
target_ids[:, :-trg_len] = -<span class="hljs-number">100</span>
<span class="hljs-keyword">with</span> torch.no_grad():
outputs = model(input_ids, labels=target_ids)
<span class="hljs-comment"># loss is calculated using CrossEntropyLoss which averages over valid labels</span>
<span class="hljs-comment"># N.B. the model only calculates loss over trg_len - 1 labels, because it internally shifts the labels</span>
<span class="hljs-comment"># to the left by 1.</span>
neg_log_likelihood = outputs.loss
<span class="hljs-comment"># Accumulate the total negative log-likelihood and the total number of tokens</span>
num_valid_tokens = (target_ids != -<span class="hljs-number">100</span>).<span class="hljs-built_in">sum</span>().item() <span class="hljs-comment"># number of valid tokens in target_ids</span>
batch_size = target_ids.size(<span class="hljs-number">0</span>)
num_loss_tokens = num_valid_tokens - batch_size <span class="hljs-comment"># subtract batch_size due to internal label shift</span>
nll_sum += neg_log_likelihood * num_loss_tokens
n_tokens += num_loss_tokens
prev_end_loc = end_loc
<span class="hljs-keyword">if</span> end_loc == seq_len:
<span class="hljs-keyword">break</span>
avg_nll = nll_sum / n_tokens <span class="hljs-comment"># average negative log-likelihood per token</span>
ppl = torch.exp(avg_nll)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-6e6tqv">Running this with the stride length equal to the max input length is equivalent to the suboptimal, non-sliding-window
strategy we discussed above. The smaller the stride, the more context the model will have in making each prediction,
and the better the reported perplexity will typically be.</p> <p data-svelte-h="svelte-13g6dq7">When we run the above with <code>stride = 1024</code>, i.e. no overlap, the resulting PPL is <code>19.44</code>, which is about the same
as the <code>19.93</code> reported in the GPT-2 paper. By using <code>stride = 512</code> and thereby employing our striding window
strategy, this jumps down to <code>16.44</code>. This is not only a more favorable score, but is calculated in a way that is
closer to the true autoregressive decomposition of a sequence likelihood.</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/perplexity.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_z647wz = {
assets: "/docs/transformers/pr_33913/en",
base: "/docs/transformers/pr_33913/en",
env: {}
};
const element = document.currentScript.parentElement;
const data = [null,null];
Promise.all([
import("/docs/transformers/pr_33913/en/_app/immutable/entry/start.b67f883f.js"),
import("/docs/transformers/pr_33913/en/_app/immutable/entry/app.e436b1f2.js")
]).then(([kit, app]) => {
kit.start(app, element, {
node_ids: [0, 390],
data,
form: null,
error: null
});
});
}
</script>

Xet Storage Details

Size:
39 kB
·
Xet hash:
e4e17af438150ee5d0bb561354820fb3caa227935577c2332842045dbb9a54dd

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