Buckets:

hf-doc-build
/

doc-dev

hf-doc-build/doc-dev / transformers /pr_36839 /en /_app /immutable /nodes /113.962d2eeb.js

rtrm's picture

4 months ago

239 kB

	import{s as Dn,o as On,n as z}from"../chunks/scheduler.01eeda35.js";import{S as Kn,i as eo,g as m,s as r,r as T,A as to,h as u,f as a,c as i,j as W,u as v,x as _,k as Z,y as l,a as d,v as M,d as y,t as k,w}from"../chunks/index.6dd51b66.js";import{T as Qe}from"../chunks/Tip.de9bae2b.js";import{D as q}from"../chunks/Docstring.76e6b3cf.js";import{C as je}from"../chunks/CodeBlock.864da1b0.js";import{F as no,M as Pn}from"../chunks/Markdown.3138439e.js";import{E as Be}from"../chunks/ExampleCodeBlock.6a36fb6b.js";import{H as Ne,E as oo}from"../chunks/EditOnGithub.7faefd25.js";function so(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> ConvBertConfig, ConvBertModel

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># Initializing a ConvBERT convbert-base-uncased style configuration</span>
	<span class="hljs-meta">>>> </span>configuration = ConvBertConfig()

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># Initializing a model (with random weights) from the convbert-base-uncased style configuration</span>
	<span class="hljs-meta">>>> </span>model = ConvBertModel(configuration)

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># Accessing the model configuration</span>
	<span class="hljs-meta">>>> </span>configuration = model.config`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function ao(C){let e,f="pair mask has the following format:",n,o,b;return o=new je({props:{code:"MCUyMDAlMjAwJTIwMCUyMDAlMjAwJTIwMCUyMDAlMjAwJTIwMCUyMDAlMjAxJTIwMSUyMDElMjAxJTIwMSUyMDElMjAxJTIwMSUyMDElMEElN0MlMjBmaXJzdCUyMHNlcXVlbmNlJTIwJTIwJTIwJTIwJTdDJTIwc2Vjb25kJTIwc2VxdWVuY2UlMjAlN0M=",highlighted:`0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 1 </span>1<span class="hljs-number"> 1 </span>1<span class="hljs-number"> 1 </span>1<span class="hljs-number"> 1 </span>1 1
	\| first sequence \| second sequence \|`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-qjgeij"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function ro(C){let e,f="pair mask has the following format:",n,o,b;return o=new je({props:{code:"MCUyMDAlMjAwJTIwMCUyMDAlMjAwJTIwMCUyMDAlMjAwJTIwMCUyMDAlMjAxJTIwMSUyMDElMjAxJTIwMSUyMDElMjAxJTIwMSUyMDElMEElN0MlMjBmaXJzdCUyMHNlcXVlbmNlJTIwJTIwJTIwJTIwJTdDJTIwc2Vjb25kJTIwc2VxdWVuY2UlMjAlN0M=",highlighted:`0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 0 </span>0<span class="hljs-number"> 1 </span>1<span class="hljs-number"> 1 </span>1<span class="hljs-number"> 1 </span>1<span class="hljs-number"> 1 </span>1 1
	\| first sequence \| second sequence \|`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-qjgeij"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function io(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function lo(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, ConvBertModel
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> torch

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = ConvBertModel.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(<span class="hljs-string">"Hello, my dog is cute"</span>, return_tensors=<span class="hljs-string">"pt"</span>)
	<span class="hljs-meta">>>> </span>outputs = model(**inputs)

	<span class="hljs-meta">>>> </span>last_hidden_states = outputs.last_hidden_state`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function co(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function po(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, ConvBertForMaskedLM
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> torch

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = ConvBertForMaskedLM.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(<span class="hljs-string">"The capital of France is [MASK]."</span>, return_tensors=<span class="hljs-string">"pt"</span>)

	<span class="hljs-meta">>>> </span><span class="hljs-keyword">with</span> torch.no_grad():
	<span class="hljs-meta">... </span> logits = model(**inputs).logits

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># retrieve index of [MASK]</span>
	<span class="hljs-meta">>>> </span>mask_token_index = (inputs.input_ids == tokenizer.mask_token_id)[<span class="hljs-number">0</span>].nonzero(as_tuple=<span class="hljs-literal">True</span>)[<span class="hljs-number">0</span>]

	<span class="hljs-meta">>>> </span>predicted_token_id = logits[<span class="hljs-number">0</span>, mask_token_index].argmax(axis=-<span class="hljs-number">1</span>)

	<span class="hljs-meta">>>> </span>labels = tokenizer(<span class="hljs-string">"The capital of France is Paris."</span>, return_tensors=<span class="hljs-string">"pt"</span>)[<span class="hljs-string">"input_ids"</span>]
	<span class="hljs-meta">>>> </span><span class="hljs-comment"># mask labels of non-[MASK] tokens</span>
	<span class="hljs-meta">>>> </span>labels = torch.where(inputs.input_ids == tokenizer.mask_token_id, labels, -<span class="hljs-number">100</span>)

	<span class="hljs-meta">>>> </span>outputs = model(**inputs, labels=labels)`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function mo(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function uo(C){let e,f="Example of single-label classification:",n,o,b;return o=new je({props:{code:"aW1wb3J0JTIwdG9yY2glMEFmcm9tJTIwdHJhbnNmb3JtZXJzJTIwaW1wb3J0JTIwQXV0b1Rva2VuaXplciUyQyUyMENvbnZCZXJ0Rm9yU2VxdWVuY2VDbGFzc2lmaWNhdGlvbiUwQSUwQXRva2VuaXplciUyMCUzRCUyMEF1dG9Ub2tlbml6ZXIuZnJvbV9wcmV0cmFpbmVkKCUyMllpdHVUZWNoJTJGY29udi1iZXJ0LWJhc2UlMjIpJTBBbW9kZWwlMjAlM0QlMjBDb252QmVydEZvclNlcXVlbmNlQ2xhc3NpZmljYXRpb24uZnJvbV9wcmV0cmFpbmVkKCUyMllpdHVUZWNoJTJGY29udi1iZXJ0LWJhc2UlMjIpJTBBJTBBaW5wdXRzJTIwJTNEJTIwdG9rZW5pemVyKCUyMkhlbGxvJTJDJTIwbXklMjBkb2clMjBpcyUyMGN1dGUlMjIlMkMlMjByZXR1cm5fdGVuc29ycyUzRCUyMnB0JTIyKSUwQSUwQXdpdGglMjB0b3JjaC5ub19ncmFkKCklM0ElMEElMjAlMjAlMjAlMjBsb2dpdHMlMjAlM0QlMjBtb2RlbCgqKmlucHV0cykubG9naXRzJTBBJTBBcHJlZGljdGVkX2NsYXNzX2lkJTIwJTNEJTIwbG9naXRzLmFyZ21heCgpLml0ZW0oKSUwQSUwQSUyMyUyMFRvJTIwdHJhaW4lMjBhJTIwbW9kZWwlMjBvbiUyMCU2MG51bV9sYWJlbHMlNjAlMjBjbGFzc2VzJTJDJTIweW91JTIwY2FuJTIwcGFzcyUyMCU2MG51bV9sYWJlbHMlM0RudW1fbGFiZWxzJTYwJTIwdG8lMjAlNjAuZnJvbV9wcmV0cmFpbmVkKC4uLiklNjAlMEFudW1fbGFiZWxzJTIwJTNEJTIwbGVuKG1vZGVsLmNvbmZpZy5pZDJsYWJlbCklMEFtb2RlbCUyMCUzRCUyMENvbnZCZXJ0Rm9yU2VxdWVuY2VDbGFzc2lmaWNhdGlvbi5mcm9tX3ByZXRyYWluZWQoJTIyWWl0dVRlY2glMkZjb252LWJlcnQtYmFzZSUyMiUyQyUyMG51bV9sYWJlbHMlM0RudW1fbGFiZWxzKSUwQSUwQWxhYmVscyUyMCUzRCUyMHRvcmNoLnRlbnNvciglNUIxJTVEKSUwQWxvc3MlMjAlM0QlMjBtb2RlbCgqKmlucHV0cyUyQyUyMGxhYmVscyUzRGxhYmVscykubG9zcw==",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> torch
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, ConvBertForSequenceClassification

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = ConvBertForSequenceClassification.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(<span class="hljs-string">"Hello, my dog is cute"</span>, return_tensors=<span class="hljs-string">"pt"</span>)

	<span class="hljs-meta">>>> </span><span class="hljs-keyword">with</span> torch.no_grad():
	<span class="hljs-meta">... </span> logits = model(**inputs).logits

	<span class="hljs-meta">>>> </span>predicted_class_id = logits.argmax().item()

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># To train a model on \`num_labels\` classes, you can pass \`num_labels=num_labels\` to \`.from_pretrained(...)\`</span>
	<span class="hljs-meta">>>> </span>num_labels = <span class="hljs-built_in">len</span>(model.config.id2label)
	<span class="hljs-meta">>>> </span>model = ConvBertForSequenceClassification.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>, num_labels=num_labels)

	<span class="hljs-meta">>>> </span>labels = torch.tensor([<span class="hljs-number">1</span>])
	<span class="hljs-meta">>>> </span>loss = model(**inputs, labels=labels).loss`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-ykxpe4"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function ho(C){let e,f="Example of multi-label classification:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> torch
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, ConvBertForSequenceClassification

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = ConvBertForSequenceClassification.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>, problem_type=<span class="hljs-string">"multi_label_classification"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(<span class="hljs-string">"Hello, my dog is cute"</span>, return_tensors=<span class="hljs-string">"pt"</span>)

	<span class="hljs-meta">>>> </span><span class="hljs-keyword">with</span> torch.no_grad():
	<span class="hljs-meta">... </span> logits = model(**inputs).logits

	<span class="hljs-meta">>>> </span>predicted_class_ids = torch.arange(<span class="hljs-number">0</span>, logits.shape[-<span class="hljs-number">1</span>])[torch.sigmoid(logits).squeeze(dim=<span class="hljs-number">0</span>) > <span class="hljs-number">0.5</span>]

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># To train a model on \`num_labels\` classes, you can pass \`num_labels=num_labels\` to \`.from_pretrained(...)\`</span>
	<span class="hljs-meta">>>> </span>num_labels = <span class="hljs-built_in">len</span>(model.config.id2label)
	<span class="hljs-meta">>>> </span>model = ConvBertForSequenceClassification.from_pretrained(
	<span class="hljs-meta">... </span> <span class="hljs-string">"YituTech/conv-bert-base"</span>, num_labels=num_labels, problem_type=<span class="hljs-string">"multi_label_classification"</span>
	<span class="hljs-meta">... </span>)

	<span class="hljs-meta">>>> </span>labels = torch.<span class="hljs-built_in">sum</span>(
	<span class="hljs-meta">... </span> torch.nn.functional.one_hot(predicted_class_ids[<span class="hljs-literal">None</span>, :].clone(), num_classes=num_labels), dim=<span class="hljs-number">1</span>
	<span class="hljs-meta">... </span>).to(torch.<span class="hljs-built_in">float</span>)
	<span class="hljs-meta">>>> </span>loss = model(**inputs, labels=labels).loss`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-1l8e32d"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function fo(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function go(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, ConvBertForMultipleChoice
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> torch

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = ConvBertForMultipleChoice.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>prompt = <span class="hljs-string">"In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."</span>
	<span class="hljs-meta">>>> </span>choice0 = <span class="hljs-string">"It is eaten with a fork and a knife."</span>
	<span class="hljs-meta">>>> </span>choice1 = <span class="hljs-string">"It is eaten while held in the hand."</span>
	<span class="hljs-meta">>>> </span>labels = torch.tensor(<span class="hljs-number">0</span>).unsqueeze(<span class="hljs-number">0</span>) <span class="hljs-comment"># choice0 is correct (according to Wikipedia ;)), batch size 1</span>

	<span class="hljs-meta">>>> </span>encoding = tokenizer([prompt, prompt], [choice0, choice1], return_tensors=<span class="hljs-string">"pt"</span>, padding=<span class="hljs-literal">True</span>)
	<span class="hljs-meta">>>> </span>outputs = model(**{k: v.unsqueeze(<span class="hljs-number">0</span>) <span class="hljs-keyword">for</span> k, v <span class="hljs-keyword">in</span> encoding.items()}, labels=labels) <span class="hljs-comment"># batch size is 1</span>

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># the linear classifier still needs to be trained</span>
	<span class="hljs-meta">>>> </span>loss = outputs.loss
	<span class="hljs-meta">>>> </span>logits = outputs.logits`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function _o(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function bo(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, ConvBertForTokenClassification
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> torch

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = ConvBertForTokenClassification.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(
	<span class="hljs-meta">... </span> <span class="hljs-string">"HuggingFace is a company based in Paris and New York"</span>, add_special_tokens=<span class="hljs-literal">False</span>, return_tensors=<span class="hljs-string">"pt"</span>
	<span class="hljs-meta">... </span>)

	<span class="hljs-meta">>>> </span><span class="hljs-keyword">with</span> torch.no_grad():
	<span class="hljs-meta">... </span> logits = model(**inputs).logits

	<span class="hljs-meta">>>> </span>predicted_token_class_ids = logits.argmax(-<span class="hljs-number">1</span>)

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># Note that tokens are classified rather then input words which means that</span>
	<span class="hljs-meta">>>> </span><span class="hljs-comment"># there might be more predicted token classes than words.</span>
	<span class="hljs-meta">>>> </span><span class="hljs-comment"># Multiple token classes might account for the same word</span>
	<span class="hljs-meta">>>> </span>predicted_tokens_classes = [model.config.id2label[t.item()] <span class="hljs-keyword">for</span> t <span class="hljs-keyword">in</span> predicted_token_class_ids[<span class="hljs-number">0</span>]]

	<span class="hljs-meta">>>> </span>labels = predicted_token_class_ids
	<span class="hljs-meta">>>> </span>loss = model(**inputs, labels=labels).loss`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function To(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function vo(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, ConvBertForQuestionAnswering
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> torch

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = ConvBertForQuestionAnswering.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>question, text = <span class="hljs-string">"Who was Jim Henson?"</span>, <span class="hljs-string">"Jim Henson was a nice puppet"</span>

	<span class="hljs-meta">>>> </span>inputs = tokenizer(question, text, return_tensors=<span class="hljs-string">"pt"</span>)
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">with</span> torch.no_grad():
	<span class="hljs-meta">... </span> outputs = model(**inputs)

	<span class="hljs-meta">>>> </span>answer_start_index = outputs.start_logits.argmax()
	<span class="hljs-meta">>>> </span>answer_end_index = outputs.end_logits.argmax()

	<span class="hljs-meta">>>> </span>predict_answer_tokens = inputs.input_ids[<span class="hljs-number">0</span>, answer_start_index : answer_end_index + <span class="hljs-number">1</span>]

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># target is "nice puppet"</span>
	<span class="hljs-meta">>>> </span>target_start_index = torch.tensor([<span class="hljs-number">14</span>])
	<span class="hljs-meta">>>> </span>target_end_index = torch.tensor([<span class="hljs-number">15</span>])

	<span class="hljs-meta">>>> </span>outputs = model(**inputs, start_positions=target_start_index, end_positions=target_end_index)
	<span class="hljs-meta">>>> </span>loss = outputs.loss`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function Mo(C){let e,f,n,o,b,t,g=`The bare ConvBERT Model transformer outputting raw hidden-states without any specific head on top.
	This model is a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> sub-class. Use
	it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
	behavior.`,E,U,J,R,x,F='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertModel">ConvBertModel</a> forward method, overrides the <code>__call__</code> special method.',V,p,j,P,Te,jt,et,fe,Jt,Ye,qe,tt=`ConvBERT Model with a <code>language modeling</code> head on top.
	This model is a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> sub-class. Use
	it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
	behavior.`,nt,le,Se,H,Je,Ft='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertForMaskedLM">ConvBertForMaskedLM</a> forward method, overrides the <code>__call__</code> special method.',Et,ot,ct,ve,_e,Fe,pt,de,Ut,st,Ae,Me=`ConvBERT Model transformer with a sequence classification/regression head on top (a linear layer on top of the
	pooled output) e.g. for GLUE tasks.`,mt,Ue,xt=`This model is a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> sub-class. Use
	it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
	behavior.`,zt,ce,I,ut,xe,Qt='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertForSequenceClassification">ConvBertForSequenceClassification</a> forward method, overrides the <code>__call__</code> special method.',ht,Pe,Re,at,K,S,rt,Ve,It,ye,Ge,Yt,St,At=`ConvBERT Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a
	softmax) e.g. for RocStories/SWAG tasks.`,ft,He,gt=`This model is a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> sub-class. Use
	it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
	behavior.`,Zt,G,ze,ke,be,pn='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertForMultipleChoice">ConvBertForMultipleChoice</a> forward method, overrides the <code>__call__</code> special method.',Wt,we,sn,D,nn,pe,on,ee,Ie,L,Ze,Nt=`ConvBERT Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
	Named-Entity-Recognition (NER) tasks.`,_t,te,qt=`This model is a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> sub-class. Use
	it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
	behavior.`,Rt,ge,Vt,Gt,it,mn='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertForTokenClassification">ConvBertForTokenClassification</a> forward method, overrides the <code>__call__</code> special method.',lt,me,oe,Ce,bt,De,an,se,We,rn,Q,Tt=`ConvBERT Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear
	layers on top of the hidden-states output to compute <code>span start logits</code> and <code>span end logits</code>).`,vt,Le,ue=`This model is a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> sub-class. Use
	it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
	behavior.`,Mt,ne,Xe,Pt,Dt,Ht='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertForQuestionAnswering">ConvBertForQuestionAnswering</a> forward method, overrides the <code>__call__</code> special method.',yt,Oe,Ee,dt,Lt;return e=new Ne({props:{title:"ConvBertModel",local:"transformers.ConvBertModel",headingTag:"h2"}}),o=new q({props:{name:"class transformers.ConvBertModel",anchor:"transformers.ConvBertModel",parameters:[{name:"config",val:""}],parametersDescription:[{anchor:"transformers.ConvBertModel.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L745"}}),J=new q({props:{name:"forward",anchor:"transformers.ConvBertModel.forward",parameters:[{name:"input_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"attention_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"token_type_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"position_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"head_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"inputs_embeds",val:": typing.Optional[torch.FloatTensor] = None"},{name:"output_attentions",val:": typing.Optional[bool] = None"},{name:"output_hidden_states",val:": typing.Optional[bool] = None"},{name:"return_dict",val:": typing.Optional[bool] = None"}],parametersDescription:[{anchor:"transformers.ConvBertModel.forward.input_ids",description:`<strong>input_ids</strong> (<code>torch.LongTensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.ConvBertModel.forward.attention_mask",description:`<strong>attention_mask</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:`,name:"attention_mask"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L776",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.BaseModelOutputWithCrossAttentions"
	>transformers.modeling_outputs.BaseModelOutputWithCrossAttentions</a> or a tuple of
	<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
	elements depending on the configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>last_hidden_state</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>) — Sequence of hidden-states at the output of the last layer of the model.</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
	one for the output of each layer) of shape <code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	<li>
	<p><strong>cross_attentions</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> and <code>config.add_cross_attention=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights of the decoder’s cross-attention layer, after the attention softmax, used to compute the
	weighted average in the cross-attention heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.BaseModelOutputWithCrossAttentions"
	>transformers.modeling_outputs.BaseModelOutputWithCrossAttentions</a> or <code>tuple(torch.FloatTensor)</code></p>
	`}}),p=new Qe({props:{$$slots:{default:[io]},$$scope:{ctx:C}}}),P=new Be({props:{anchor:"transformers.ConvBertModel.forward.example",$$slots:{default:[lo]},$$scope:{ctx:C}}}),jt=new Ne({props:{title:"ConvBertForMaskedLM",local:"transformers.ConvBertForMaskedLM",headingTag:"h2"}}),Jt=new q({props:{name:"class transformers.ConvBertForMaskedLM",anchor:"transformers.ConvBertForMaskedLM",parameters:[{name:"config",val:""}],parametersDescription:[{anchor:"transformers.ConvBertForMaskedLM.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L863"}}),Se=new q({props:{name:"forward",anchor:"transformers.ConvBertForMaskedLM.forward",parameters:[{name:"input_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"attention_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"token_type_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"position_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"head_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"inputs_embeds",val:": typing.Optional[torch.FloatTensor] = None"},{name:"labels",val:": typing.Optional[torch.LongTensor] = None"},{name:"output_attentions",val:": typing.Optional[bool] = None"},{name:"output_hidden_states",val:": typing.Optional[bool] = None"},{name:"return_dict",val:": typing.Optional[bool] = None"}],parametersDescription:[{anchor:"transformers.ConvBertForMaskedLM.forward.input_ids",description:`<strong>input_ids</strong> (<code>torch.LongTensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.ConvBertForMaskedLM.forward.attention_mask",description:`<strong>attention_mask</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:`,name:"attention_mask"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L883",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.MaskedLMOutput"
	>transformers.modeling_outputs.MaskedLMOutput</a> or a tuple of
	<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
	elements depending on the configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>torch.FloatTensor</code> of shape <code>(1,)</code>, <em>optional</em>, returned when <code>labels</code> is provided) — Masked language modeling (MLM) loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length, config.vocab_size)</code>) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
	one for the output of each layer) of shape <code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.MaskedLMOutput"
	>transformers.modeling_outputs.MaskedLMOutput</a> or <code>tuple(torch.FloatTensor)</code></p>
	`}}),ot=new Qe({props:{$$slots:{default:[co]},$$scope:{ctx:C}}}),ve=new Be({props:{anchor:"transformers.ConvBertForMaskedLM.forward.example",$$slots:{default:[po]},$$scope:{ctx:C}}}),Fe=new Ne({props:{title:"ConvBertForSequenceClassification",local:"transformers.ConvBertForSequenceClassification",headingTag:"h2"}}),Ut=new q({props:{name:"class transformers.ConvBertForSequenceClassification",anchor:"transformers.ConvBertForSequenceClassification",parameters:[{name:"config",val:""}],parametersDescription:[{anchor:"transformers.ConvBertForSequenceClassification.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L968"}}),I=new q({props:{name:"forward",anchor:"transformers.ConvBertForSequenceClassification.forward",parameters:[{name:"input_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"attention_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"token_type_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"position_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"head_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"inputs_embeds",val:": typing.Optional[torch.FloatTensor] = None"},{name:"labels",val:": typing.Optional[torch.LongTensor] = None"},{name:"output_attentions",val:": typing.Optional[bool] = None"},{name:"output_hidden_states",val:": typing.Optional[bool] = None"},{name:"return_dict",val:": typing.Optional[bool] = None"}],parametersDescription:[{anchor:"transformers.ConvBertForSequenceClassification.forward.input_ids",description:`<strong>input_ids</strong> (<code>torch.LongTensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.ConvBertForSequenceClassification.forward.attention_mask",description:`<strong>attention_mask</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:`,name:"attention_mask"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L986",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.SequenceClassifierOutput"
	>transformers.modeling_outputs.SequenceClassifierOutput</a> or a tuple of
	<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
	elements depending on the configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>torch.FloatTensor</code> of shape <code>(1,)</code>, <em>optional</em>, returned when <code>labels</code> is provided) — Classification (or regression if config.num_labels==1) loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, config.num_labels)</code>) — Classification (or regression if config.num_labels==1) scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
	one for the output of each layer) of shape <code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.SequenceClassifierOutput"
	>transformers.modeling_outputs.SequenceClassifierOutput</a> or <code>tuple(torch.FloatTensor)</code></p>
	`}}),Pe=new Qe({props:{$$slots:{default:[mo]},$$scope:{ctx:C}}}),at=new Be({props:{anchor:"transformers.ConvBertForSequenceClassification.forward.example",$$slots:{default:[uo]},$$scope:{ctx:C}}}),S=new Be({props:{anchor:"transformers.ConvBertForSequenceClassification.forward.example-2",$$slots:{default:[ho]},$$scope:{ctx:C}}}),Ve=new Ne({props:{title:"ConvBertForMultipleChoice",local:"transformers.ConvBertForMultipleChoice",headingTag:"h2"}}),Ge=new q({props:{name:"class transformers.ConvBertForMultipleChoice",anchor:"transformers.ConvBertForMultipleChoice",parameters:[{name:"config",val:""}],parametersDescription:[{anchor:"transformers.ConvBertForMultipleChoice.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L1063"}}),ze=new q({props:{name:"forward",anchor:"transformers.ConvBertForMultipleChoice.forward",parameters:[{name:"input_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"attention_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"token_type_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"position_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"head_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"inputs_embeds",val:": typing.Optional[torch.FloatTensor] = None"},{name:"labels",val:": typing.Optional[torch.LongTensor] = None"},{name:"output_attentions",val:": typing.Optional[bool] = None"},{name:"output_hidden_states",val:": typing.Optional[bool] = None"},{name:"return_dict",val:": typing.Optional[bool] = None"}],parametersDescription:[{anchor:"transformers.ConvBertForMultipleChoice.forward.input_ids",description:`<strong>input_ids</strong> (<code>torch.LongTensor</code> of shape <code>(batch_size, num_choices, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.ConvBertForMultipleChoice.forward.attention_mask",description:`<strong>attention_mask</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, num_choices, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:`,name:"attention_mask"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L1081",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.MultipleChoiceModelOutput"
	>transformers.modeling_outputs.MultipleChoiceModelOutput</a> or a tuple of
	<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
	elements depending on the configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>torch.FloatTensor</code> of shape <em>(1,)</em>, <em>optional</em>, returned when <code>labels</code> is provided) — Classification loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, num_choices)</code>) — <em>num_choices</em> is the second dimension of the input tensors. (see <em>input_ids</em> above).</p>
	<p>Classification scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
	one for the output of each layer) of shape <code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.MultipleChoiceModelOutput"
	>transformers.modeling_outputs.MultipleChoiceModelOutput</a> or <code>tuple(torch.FloatTensor)</code></p>
	`}}),we=new Qe({props:{$$slots:{default:[fo]},$$scope:{ctx:C}}}),D=new Be({props:{anchor:"transformers.ConvBertForMultipleChoice.forward.example",$$slots:{default:[go]},$$scope:{ctx:C}}}),pe=new Ne({props:{title:"ConvBertForTokenClassification",local:"transformers.ConvBertForTokenClassification",headingTag:"h2"}}),Ie=new q({props:{name:"class transformers.ConvBertForTokenClassification",anchor:"transformers.ConvBertForTokenClassification",parameters:[{name:"config",val:""}],parametersDescription:[{anchor:"transformers.ConvBertForTokenClassification.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L1156"}}),Vt=new q({props:{name:"forward",anchor:"transformers.ConvBertForTokenClassification.forward",parameters:[{name:"input_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"attention_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"token_type_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"position_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"head_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"inputs_embeds",val:": typing.Optional[torch.FloatTensor] = None"},{name:"labels",val:": typing.Optional[torch.LongTensor] = None"},{name:"output_attentions",val:": typing.Optional[bool] = None"},{name:"output_hidden_states",val:": typing.Optional[bool] = None"},{name:"return_dict",val:": typing.Optional[bool] = None"}],parametersDescription:[{anchor:"transformers.ConvBertForTokenClassification.forward.input_ids",description:`<strong>input_ids</strong> (<code>torch.LongTensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.ConvBertForTokenClassification.forward.attention_mask",description:`<strong>attention_mask</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:`,name:"attention_mask"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L1178",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.TokenClassifierOutput"
	>transformers.modeling_outputs.TokenClassifierOutput</a> or a tuple of
	<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
	elements depending on the configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>torch.FloatTensor</code> of shape <code>(1,)</code>, <em>optional</em>, returned when <code>labels</code> is provided) — Classification loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length, config.num_labels)</code>) — Classification scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
	one for the output of each layer) of shape <code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.TokenClassifierOutput"
	>transformers.modeling_outputs.TokenClassifierOutput</a> or <code>tuple(torch.FloatTensor)</code></p>
	`}}),me=new Qe({props:{$$slots:{default:[_o]},$$scope:{ctx:C}}}),Ce=new Be({props:{anchor:"transformers.ConvBertForTokenClassification.forward.example",$$slots:{default:[bo]},$$scope:{ctx:C}}}),De=new Ne({props:{title:"ConvBertForQuestionAnswering",local:"transformers.ConvBertForQuestionAnswering",headingTag:"h2"}}),We=new q({props:{name:"class transformers.ConvBertForQuestionAnswering",anchor:"transformers.ConvBertForQuestionAnswering",parameters:[{name:"config",val:""}],parametersDescription:[{anchor:"transformers.ConvBertForQuestionAnswering.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L1237"}}),Xe=new q({props:{name:"forward",anchor:"transformers.ConvBertForQuestionAnswering.forward",parameters:[{name:"input_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"attention_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"token_type_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"position_ids",val:": typing.Optional[torch.LongTensor] = None"},{name:"head_mask",val:": typing.Optional[torch.FloatTensor] = None"},{name:"inputs_embeds",val:": typing.Optional[torch.FloatTensor] = None"},{name:"start_positions",val:": typing.Optional[torch.LongTensor] = None"},{name:"end_positions",val:": typing.Optional[torch.LongTensor] = None"},{name:"output_attentions",val:": typing.Optional[bool] = None"},{name:"output_hidden_states",val:": typing.Optional[bool] = None"},{name:"return_dict",val:": typing.Optional[bool] = None"}],parametersDescription:[{anchor:"transformers.ConvBertForQuestionAnswering.forward.input_ids",description:`<strong>input_ids</strong> (<code>torch.LongTensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.ConvBertForQuestionAnswering.forward.attention_mask",description:`<strong>attention_mask</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:`,name:"attention_mask"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_convbert.py#L1255",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.QuestionAnsweringModelOutput"
	>transformers.modeling_outputs.QuestionAnsweringModelOutput</a> or a tuple of
	<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
	elements depending on the configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>torch.FloatTensor</code> of shape <code>(1,)</code>, <em>optional</em>, returned when <code>labels</code> is provided) — Total span extraction loss is the sum of a Cross-Entropy for the start and end positions.</p>
	</li>
	<li>
	<p><strong>start_logits</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length)</code>) — Span-start scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>end_logits</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length)</code>) — Span-end scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
	one for the output of each layer) of shape <code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_outputs.QuestionAnsweringModelOutput"
	>transformers.modeling_outputs.QuestionAnsweringModelOutput</a> or <code>tuple(torch.FloatTensor)</code></p>
	`}}),Oe=new Qe({props:{$$slots:{default:[To]},$$scope:{ctx:C}}}),dt=new Be({props:{anchor:"transformers.ConvBertForQuestionAnswering.forward.example",$$slots:{default:[vo]},$$scope:{ctx:C}}}),{c(){T(e.$$.fragment),f=r(),n=m("div"),T(o.$$.fragment),b=r(),t=m("p"),t.innerHTML=g,E=r(),U=m("div"),T(J.$$.fragment),R=r(),x=m("p"),x.innerHTML=F,V=r(),T(p.$$.fragment),j=r(),T(P.$$.fragment),Te=r(),T(jt.$$.fragment),et=r(),fe=m("div"),T(Jt.$$.fragment),Ye=r(),qe=m("p"),qe.innerHTML=tt,nt=r(),le=m("div"),T(Se.$$.fragment),H=r(),Je=m("p"),Je.innerHTML=Ft,Et=r(),T(ot.$$.fragment),ct=r(),T(ve.$$.fragment),_e=r(),T(Fe.$$.fragment),pt=r(),de=m("div"),T(Ut.$$.fragment),st=r(),Ae=m("p"),Ae.textContent=Me,mt=r(),Ue=m("p"),Ue.innerHTML=xt,zt=r(),ce=m("div"),T(I.$$.fragment),ut=r(),xe=m("p"),xe.innerHTML=Qt,ht=r(),T(Pe.$$.fragment),Re=r(),T(at.$$.fragment),K=r(),T(S.$$.fragment),rt=r(),T(Ve.$$.fragment),It=r(),ye=m("div"),T(Ge.$$.fragment),Yt=r(),St=m("p"),St.textContent=At,ft=r(),He=m("p"),He.innerHTML=gt,Zt=r(),G=m("div"),T(ze.$$.fragment),ke=r(),be=m("p"),be.innerHTML=pn,Wt=r(),T(we.$$.fragment),sn=r(),T(D.$$.fragment),nn=r(),T(pe.$$.fragment),on=r(),ee=m("div"),T(Ie.$$.fragment),L=r(),Ze=m("p"),Ze.textContent=Nt,_t=r(),te=m("p"),te.innerHTML=qt,Rt=r(),ge=m("div"),T(Vt.$$.fragment),Gt=r(),it=m("p"),it.innerHTML=mn,lt=r(),T(me.$$.fragment),oe=r(),T(Ce.$$.fragment),bt=r(),T(De.$$.fragment),an=r(),se=m("div"),T(We.$$.fragment),rn=r(),Q=m("p"),Q.innerHTML=Tt,vt=r(),Le=m("p"),Le.innerHTML=ue,Mt=r(),ne=m("div"),T(Xe.$$.fragment),Pt=r(),Dt=m("p"),Dt.innerHTML=Ht,yt=r(),T(Oe.$$.fragment),Ee=r(),T(dt.$$.fragment),this.h()},l(h){v(e.$$.fragment,h),f=i(h),n=u(h,"DIV",{class:!0});var s=W(n);v(o.$$.fragment,s),b=i(s),t=u(s,"P",{"data-svelte-h":!0}),_(t)!=="svelte-1v79ih5"&&(t.innerHTML=g),E=i(s),U=u(s,"DIV",{class:!0});var $=W(U);v(J.$$.fragment,$),R=i($),x=u($,"P",{"data-svelte-h":!0}),_(x)!=="svelte-yfn7uh"&&(x.innerHTML=F),V=i($),v(p.$$.fragment,$),j=i($),v(P.$$.fragment,$),$.forEach(a),s.forEach(a),Te=i(h),v(jt.$$.fragment,h),et=i(h),fe=u(h,"DIV",{class:!0});var X=W(fe);v(Jt.$$.fragment,X),Ye=i(X),qe=u(X,"P",{"data-svelte-h":!0}),_(qe)!=="svelte-16ubxu4"&&(qe.innerHTML=tt),nt=i(X),le=u(X,"DIV",{class:!0});var N=W(le);v(Se.$$.fragment,N),H=i(N),Je=u(N,"P",{"data-svelte-h":!0}),_(Je)!=="svelte-1ra1nuh"&&(Je.innerHTML=Ft),Et=i(N),v(ot.$$.fragment,N),ct=i(N),v(ve.$$.fragment,N),N.forEach(a),X.forEach(a),_e=i(h),v(Fe.$$.fragment,h),pt=i(h),de=u(h,"DIV",{class:!0});var Y=W(de);v(Ut.$$.fragment,Y),st=i(Y),Ae=u(Y,"P",{"data-svelte-h":!0}),_(Ae)!=="svelte-g7aq4t"&&(Ae.textContent=Me),mt=i(Y),Ue=u(Y,"P",{"data-svelte-h":!0}),_(Ue)!=="svelte-68lg8f"&&(Ue.innerHTML=xt),zt=i(Y),ce=u(Y,"DIV",{class:!0});var ae=W(ce);v(I.$$.fragment,ae),ut=i(ae),xe=u(ae,"P",{"data-svelte-h":!0}),_(xe)!=="svelte-qhq7xd"&&(xe.innerHTML=Qt),ht=i(ae),v(Pe.$$.fragment,ae),Re=i(ae),v(at.$$.fragment,ae),K=i(ae),v(S.$$.fragment,ae),ae.forEach(a),Y.forEach(a),rt=i(h),v(Ve.$$.fragment,h),It=i(h),ye=u(h,"DIV",{class:!0});var re=W(ye);v(Ge.$$.fragment,re),Yt=i(re),St=u(re,"P",{"data-svelte-h":!0}),_(St)!=="svelte-1xscdal"&&(St.textContent=At),ft=i(re),He=u(re,"P",{"data-svelte-h":!0}),_(He)!=="svelte-68lg8f"&&(He.innerHTML=gt),Zt=i(re),G=u(re,"DIV",{class:!0});var he=W(G);v(ze.$$.fragment,he),ke=i(he),be=u(he,"P",{"data-svelte-h":!0}),_(be)!=="svelte-7v7x15"&&(be.innerHTML=pn),Wt=i(he),v(we.$$.fragment,he),sn=i(he),v(D.$$.fragment,he),he.forEach(a),re.forEach(a),nn=i(h),v(pe.$$.fragment,h),on=i(h),ee=u(h,"DIV",{class:!0});var O=W(ee);v(Ie.$$.fragment,O),L=i(O),Ze=u(O,"P",{"data-svelte-h":!0}),_(Ze)!=="svelte-6su72q"&&(Ze.textContent=Nt),_t=i(O),te=u(O,"P",{"data-svelte-h":!0}),_(te)!=="svelte-68lg8f"&&(te.innerHTML=qt),Rt=i(O),ge=u(O,"DIV",{class:!0});var A=W(ge);v(Vt.$$.fragment,A),Gt=i(A),it=u(A,"P",{"data-svelte-h":!0}),_(it)!=="svelte-11ytlir"&&(it.innerHTML=mn),lt=i(A),v(me.$$.fragment,A),oe=i(A),v(Ce.$$.fragment,A),A.forEach(a),O.forEach(a),bt=i(h),v(De.$$.fragment,h),an=i(h),se=u(h,"DIV",{class:!0});var ie=W(se);v(We.$$.fragment,ie),rn=i(ie),Q=u(ie,"P",{"data-svelte-h":!0}),_(Q)!=="svelte-smf15g"&&(Q.innerHTML=Tt),vt=i(ie),Le=u(ie,"P",{"data-svelte-h":!0}),_(Le)!=="svelte-68lg8f"&&(Le.innerHTML=ue),Mt=i(ie),ne=u(ie,"DIV",{class:!0});var $e=W(ne);v(Xe.$$.fragment,$e),Pt=i($e),Dt=u($e,"P",{"data-svelte-h":!0}),_(Dt)!=="svelte-1eelwyd"&&(Dt.innerHTML=Ht),yt=i($e),v(Oe.$$.fragment,$e),Ee=i($e),v(dt.$$.fragment,$e),$e.forEach(a),ie.forEach(a),this.h()},h(){Z(U,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(n,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(le,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(fe,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ce,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(de,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(G,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ye,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ge,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ee,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ne,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(se,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8")},m(h,s){M(e,h,s),d(h,f,s),d(h,n,s),M(o,n,null),l(n,b),l(n,t),l(n,E),l(n,U),M(J,U,null),l(U,R),l(U,x),l(U,V),M(p,U,null),l(U,j),M(P,U,null),d(h,Te,s),M(jt,h,s),d(h,et,s),d(h,fe,s),M(Jt,fe,null),l(fe,Ye),l(fe,qe),l(fe,nt),l(fe,le),M(Se,le,null),l(le,H),l(le,Je),l(le,Et),M(ot,le,null),l(le,ct),M(ve,le,null),d(h,_e,s),M(Fe,h,s),d(h,pt,s),d(h,de,s),M(Ut,de,null),l(de,st),l(de,Ae),l(de,mt),l(de,Ue),l(de,zt),l(de,ce),M(I,ce,null),l(ce,ut),l(ce,xe),l(ce,ht),M(Pe,ce,null),l(ce,Re),M(at,ce,null),l(ce,K),M(S,ce,null),d(h,rt,s),M(Ve,h,s),d(h,It,s),d(h,ye,s),M(Ge,ye,null),l(ye,Yt),l(ye,St),l(ye,ft),l(ye,He),l(ye,Zt),l(ye,G),M(ze,G,null),l(G,ke),l(G,be),l(G,Wt),M(we,G,null),l(G,sn),M(D,G,null),d(h,nn,s),M(pe,h,s),d(h,on,s),d(h,ee,s),M(Ie,ee,null),l(ee,L),l(ee,Ze),l(ee,_t),l(ee,te),l(ee,Rt),l(ee,ge),M(Vt,ge,null),l(ge,Gt),l(ge,it),l(ge,lt),M(me,ge,null),l(ge,oe),M(Ce,ge,null),d(h,bt,s),M(De,h,s),d(h,an,s),d(h,se,s),M(We,se,null),l(se,rn),l(se,Q),l(se,vt),l(se,Le),l(se,Mt),l(se,ne),M(Xe,ne,null),l(ne,Pt),l(ne,Dt),l(ne,yt),M(Oe,ne,null),l(ne,Ee),M(dt,ne,null),Lt=!0},p(h,s){const $={};s&2&&($.$$scope={dirty:s,ctx:h}),p.$set($);const X={};s&2&&(X.$$scope={dirty:s,ctx:h}),P.$set(X);const N={};s&2&&(N.$$scope={dirty:s,ctx:h}),ot.$set(N);const Y={};s&2&&(Y.$$scope={dirty:s,ctx:h}),ve.$set(Y);const ae={};s&2&&(ae.$$scope={dirty:s,ctx:h}),Pe.$set(ae);const re={};s&2&&(re.$$scope={dirty:s,ctx:h}),at.$set(re);const he={};s&2&&(he.$$scope={dirty:s,ctx:h}),S.$set(he);const O={};s&2&&(O.$$scope={dirty:s,ctx:h}),we.$set(O);const A={};s&2&&(A.$$scope={dirty:s,ctx:h}),D.$set(A);const ie={};s&2&&(ie.$$scope={dirty:s,ctx:h}),me.$set(ie);const $e={};s&2&&($e.$$scope={dirty:s,ctx:h}),Ce.$set($e);const _n={};s&2&&(_n.$$scope={dirty:s,ctx:h}),Oe.$set(_n);const cn={};s&2&&(cn.$$scope={dirty:s,ctx:h}),dt.$set(cn)},i(h){Lt\|\|(y(e.$$.fragment,h),y(o.$$.fragment,h),y(J.$$.fragment,h),y(p.$$.fragment,h),y(P.$$.fragment,h),y(jt.$$.fragment,h),y(Jt.$$.fragment,h),y(Se.$$.fragment,h),y(ot.$$.fragment,h),y(ve.$$.fragment,h),y(Fe.$$.fragment,h),y(Ut.$$.fragment,h),y(I.$$.fragment,h),y(Pe.$$.fragment,h),y(at.$$.fragment,h),y(S.$$.fragment,h),y(Ve.$$.fragment,h),y(Ge.$$.fragment,h),y(ze.$$.fragment,h),y(we.$$.fragment,h),y(D.$$.fragment,h),y(pe.$$.fragment,h),y(Ie.$$.fragment,h),y(Vt.$$.fragment,h),y(me.$$.fragment,h),y(Ce.$$.fragment,h),y(De.$$.fragment,h),y(We.$$.fragment,h),y(Xe.$$.fragment,h),y(Oe.$$.fragment,h),y(dt.$$.fragment,h),Lt=!0)},o(h){k(e.$$.fragment,h),k(o.$$.fragment,h),k(J.$$.fragment,h),k(p.$$.fragment,h),k(P.$$.fragment,h),k(jt.$$.fragment,h),k(Jt.$$.fragment,h),k(Se.$$.fragment,h),k(ot.$$.fragment,h),k(ve.$$.fragment,h),k(Fe.$$.fragment,h),k(Ut.$$.fragment,h),k(I.$$.fragment,h),k(Pe.$$.fragment,h),k(at.$$.fragment,h),k(S.$$.fragment,h),k(Ve.$$.fragment,h),k(Ge.$$.fragment,h),k(ze.$$.fragment,h),k(we.$$.fragment,h),k(D.$$.fragment,h),k(pe.$$.fragment,h),k(Ie.$$.fragment,h),k(Vt.$$.fragment,h),k(me.$$.fragment,h),k(Ce.$$.fragment,h),k(De.$$.fragment,h),k(We.$$.fragment,h),k(Xe.$$.fragment,h),k(Oe.$$.fragment,h),k(dt.$$.fragment,h),Lt=!1},d(h){h&&(a(f),a(n),a(Te),a(et),a(fe),a(_e),a(pt),a(de),a(rt),a(It),a(ye),a(nn),a(on),a(ee),a(bt),a(an),a(se)),w(e,h),w(o),w(J),w(p),w(P),w(jt,h),w(Jt),w(Se),w(ot),w(ve),w(Fe,h),w(Ut),w(I),w(Pe),w(at),w(S),w(Ve,h),w(Ge),w(ze),w(we),w(D),w(pe,h),w(Ie),w(Vt),w(me),w(Ce),w(De,h),w(We),w(Xe),w(Oe),w(dt)}}}function yo(C){let e,f;return e=new Pn({props:{$$slots:{default:[Mo]},$$scope:{ctx:C}}}),{c(){T(e.$$.fragment)},l(n){v(e.$$.fragment,n)},m(n,o){M(e,n,o),f=!0},p(n,o){const b={};o&2&&(b.$$scope={dirty:o,ctx:n}),e.$set(b)},i(n){f\|\|(y(e.$$.fragment,n),f=!0)},o(n){k(e.$$.fragment,n),f=!1},d(n){w(e,n)}}}function ko(C){let e,f="TensorFlow models and layers in <code>transformers</code> accept two formats as input:",n,o,b="<li>having all inputs as keyword arguments (like PyTorch models), or</li> <li>having all inputs as a list, tuple or dict in the first positional argument.</li>",t,g,E=`The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
	and layers. Because of this support, when using methods like <code>model.fit()</code> things should “just work” for you - just
	pass your inputs and labels in any format that <code>model.fit()</code> supports! If, however, you want to use the second
	format outside of Keras methods like <code>fit()</code> and <code>predict()</code>, such as when creating your own layers or models with
	the Keras <code>Functional</code> API, there are three possibilities you can use to gather all the input Tensors in the first
	positional argument:`,U,J,R=`<li>a single Tensor with <code>input_ids</code> only and nothing else: <code>model(input_ids)</code></li> <li>a list of varying length with one or several input Tensors IN THE ORDER given in the docstring:
	<code>model([input_ids, attention_mask])</code> or <code>model([input_ids, attention_mask, token_type_ids])</code></li> <li>a dictionary with one or several input Tensors associated to the input names given in the docstring:
	<code>model({"input_ids": input_ids, "token_type_ids": token_type_ids})</code></li>`,x,F,V=`Note that when creating models and layers with
	<a href="https://keras.io/guides/making_new_layers_and_models_via_subclassing/" rel="nofollow">subclassing</a> then you don’t need to worry
	about any of this, as you can just pass inputs like you would to any other Python function!`;return{c(){e=m("p"),e.innerHTML=f,n=r(),o=m("ul"),o.innerHTML=b,t=r(),g=m("p"),g.innerHTML=E,U=r(),J=m("ul"),J.innerHTML=R,x=r(),F=m("p"),F.innerHTML=V},l(p){e=u(p,"P",{"data-svelte-h":!0}),_(e)!=="svelte-1ajbfxg"&&(e.innerHTML=f),n=i(p),o=u(p,"UL",{"data-svelte-h":!0}),_(o)!=="svelte-qm1t26"&&(o.innerHTML=b),t=i(p),g=u(p,"P",{"data-svelte-h":!0}),_(g)!=="svelte-1v9qsc5"&&(g.innerHTML=E),U=i(p),J=u(p,"UL",{"data-svelte-h":!0}),_(J)!=="svelte-15scerc"&&(J.innerHTML=R),x=i(p),F=u(p,"P",{"data-svelte-h":!0}),_(F)!=="svelte-1an3odd"&&(F.innerHTML=V)},m(p,j){d(p,e,j),d(p,n,j),d(p,o,j),d(p,t,j),d(p,g,j),d(p,U,j),d(p,J,j),d(p,x,j),d(p,F,j)},p:z,d(p){p&&(a(e),a(n),a(o),a(t),a(g),a(U),a(J),a(x),a(F))}}}function wo(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function Co(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, TFConvBertModel
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> tensorflow <span class="hljs-keyword">as</span> tf

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = TFConvBertModel.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(<span class="hljs-string">"Hello, my dog is cute"</span>, return_tensors=<span class="hljs-string">"tf"</span>)
	<span class="hljs-meta">>>> </span>outputs = model(inputs)

	<span class="hljs-meta">>>> </span>last_hidden_states = outputs.last_hidden_state`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function $o(C){let e,f="TensorFlow models and layers in <code>transformers</code> accept two formats as input:",n,o,b="<li>having all inputs as keyword arguments (like PyTorch models), or</li> <li>having all inputs as a list, tuple or dict in the first positional argument.</li>",t,g,E=`The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
	and layers. Because of this support, when using methods like <code>model.fit()</code> things should “just work” for you - just
	pass your inputs and labels in any format that <code>model.fit()</code> supports! If, however, you want to use the second
	format outside of Keras methods like <code>fit()</code> and <code>predict()</code>, such as when creating your own layers or models with
	the Keras <code>Functional</code> API, there are three possibilities you can use to gather all the input Tensors in the first
	positional argument:`,U,J,R=`<li>a single Tensor with <code>input_ids</code> only and nothing else: <code>model(input_ids)</code></li> <li>a list of varying length with one or several input Tensors IN THE ORDER given in the docstring:
	<code>model([input_ids, attention_mask])</code> or <code>model([input_ids, attention_mask, token_type_ids])</code></li> <li>a dictionary with one or several input Tensors associated to the input names given in the docstring:
	<code>model({"input_ids": input_ids, "token_type_ids": token_type_ids})</code></li>`,x,F,V=`Note that when creating models and layers with
	<a href="https://keras.io/guides/making_new_layers_and_models_via_subclassing/" rel="nofollow">subclassing</a> then you don’t need to worry
	about any of this, as you can just pass inputs like you would to any other Python function!`;return{c(){e=m("p"),e.innerHTML=f,n=r(),o=m("ul"),o.innerHTML=b,t=r(),g=m("p"),g.innerHTML=E,U=r(),J=m("ul"),J.innerHTML=R,x=r(),F=m("p"),F.innerHTML=V},l(p){e=u(p,"P",{"data-svelte-h":!0}),_(e)!=="svelte-1ajbfxg"&&(e.innerHTML=f),n=i(p),o=u(p,"UL",{"data-svelte-h":!0}),_(o)!=="svelte-qm1t26"&&(o.innerHTML=b),t=i(p),g=u(p,"P",{"data-svelte-h":!0}),_(g)!=="svelte-1v9qsc5"&&(g.innerHTML=E),U=i(p),J=u(p,"UL",{"data-svelte-h":!0}),_(J)!=="svelte-15scerc"&&(J.innerHTML=R),x=i(p),F=u(p,"P",{"data-svelte-h":!0}),_(F)!=="svelte-1an3odd"&&(F.innerHTML=V)},m(p,j){d(p,e,j),d(p,n,j),d(p,o,j),d(p,t,j),d(p,g,j),d(p,U,j),d(p,J,j),d(p,x,j),d(p,F,j)},p:z,d(p){p&&(a(e),a(n),a(o),a(t),a(g),a(U),a(J),a(x),a(F))}}}function Bo(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function jo(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, TFConvBertForMaskedLM
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> tensorflow <span class="hljs-keyword">as</span> tf

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = TFConvBertForMaskedLM.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(<span class="hljs-string">"The capital of France is [MASK]."</span>, return_tensors=<span class="hljs-string">"tf"</span>)
	<span class="hljs-meta">>>> </span>logits = model(**inputs).logits

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># retrieve index of [MASK]</span>
	<span class="hljs-meta">>>> </span>mask_token_index = tf.where((inputs.input_ids == tokenizer.mask_token_id)[<span class="hljs-number">0</span>])
	<span class="hljs-meta">>>> </span>selected_logits = tf.gather_nd(logits[<span class="hljs-number">0</span>], indices=mask_token_index)

	<span class="hljs-meta">>>> </span>predicted_token_id = tf.math.argmax(selected_logits, axis=-<span class="hljs-number">1</span>)`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function Jo(C){let e,f;return e=new je({props:{code:"bGFiZWxzJTIwJTNEJTIwdG9rZW5pemVyKCUyMlRoZSUyMGNhcGl0YWwlMjBvZiUyMEZyYW5jZSUyMGlzJTIwUGFyaXMuJTIyJTJDJTIwcmV0dXJuX3RlbnNvcnMlM0QlMjJ0ZiUyMiklNUIlMjJpbnB1dF9pZHMlMjIlNUQlMEElMjMlMjBtYXNrJTIwbGFiZWxzJTIwb2YlMjBub24tJTVCTUFTSyU1RCUyMHRva2VucyUwQWxhYmVscyUyMCUzRCUyMHRmLndoZXJlKGlucHV0cy5pbnB1dF9pZHMlMjAlM0QlM0QlMjB0b2tlbml6ZXIubWFza190b2tlbl9pZCUyQyUyMGxhYmVscyUyQyUyMC0xMDApJTBBJTBBb3V0cHV0cyUyMCUzRCUyMG1vZGVsKCoqaW5wdXRzJTJDJTIwbGFiZWxzJTNEbGFiZWxzKQ==",highlighted:`<span class="hljs-meta">>>> </span>labels = tokenizer(<span class="hljs-string">"The capital of France is Paris."</span>, return_tensors=<span class="hljs-string">"tf"</span>)[<span class="hljs-string">"input_ids"</span>]
	<span class="hljs-meta">>>> </span><span class="hljs-comment"># mask labels of non-[MASK] tokens</span>
	<span class="hljs-meta">>>> </span>labels = tf.where(inputs.input_ids == tokenizer.mask_token_id, labels, -<span class="hljs-number">100</span>)

	<span class="hljs-meta">>>> </span>outputs = model(**inputs, labels=labels)`,wrap:!1}}),{c(){T(e.$$.fragment)},l(n){v(e.$$.fragment,n)},m(n,o){M(e,n,o),f=!0},p:z,i(n){f\|\|(y(e.$$.fragment,n),f=!0)},o(n){k(e.$$.fragment,n),f=!1},d(n){w(e,n)}}}function Fo(C){let e,f="TensorFlow models and layers in <code>transformers</code> accept two formats as input:",n,o,b="<li>having all inputs as keyword arguments (like PyTorch models), or</li> <li>having all inputs as a list, tuple or dict in the first positional argument.</li>",t,g,E=`The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
	and layers. Because of this support, when using methods like <code>model.fit()</code> things should “just work” for you - just
	pass your inputs and labels in any format that <code>model.fit()</code> supports! If, however, you want to use the second
	format outside of Keras methods like <code>fit()</code> and <code>predict()</code>, such as when creating your own layers or models with
	the Keras <code>Functional</code> API, there are three possibilities you can use to gather all the input Tensors in the first
	positional argument:`,U,J,R=`<li>a single Tensor with <code>input_ids</code> only and nothing else: <code>model(input_ids)</code></li> <li>a list of varying length with one or several input Tensors IN THE ORDER given in the docstring:
	<code>model([input_ids, attention_mask])</code> or <code>model([input_ids, attention_mask, token_type_ids])</code></li> <li>a dictionary with one or several input Tensors associated to the input names given in the docstring:
	<code>model({"input_ids": input_ids, "token_type_ids": token_type_ids})</code></li>`,x,F,V=`Note that when creating models and layers with
	<a href="https://keras.io/guides/making_new_layers_and_models_via_subclassing/" rel="nofollow">subclassing</a> then you don’t need to worry
	about any of this, as you can just pass inputs like you would to any other Python function!`;return{c(){e=m("p"),e.innerHTML=f,n=r(),o=m("ul"),o.innerHTML=b,t=r(),g=m("p"),g.innerHTML=E,U=r(),J=m("ul"),J.innerHTML=R,x=r(),F=m("p"),F.innerHTML=V},l(p){e=u(p,"P",{"data-svelte-h":!0}),_(e)!=="svelte-1ajbfxg"&&(e.innerHTML=f),n=i(p),o=u(p,"UL",{"data-svelte-h":!0}),_(o)!=="svelte-qm1t26"&&(o.innerHTML=b),t=i(p),g=u(p,"P",{"data-svelte-h":!0}),_(g)!=="svelte-1v9qsc5"&&(g.innerHTML=E),U=i(p),J=u(p,"UL",{"data-svelte-h":!0}),_(J)!=="svelte-15scerc"&&(J.innerHTML=R),x=i(p),F=u(p,"P",{"data-svelte-h":!0}),_(F)!=="svelte-1an3odd"&&(F.innerHTML=V)},m(p,j){d(p,e,j),d(p,n,j),d(p,o,j),d(p,t,j),d(p,g,j),d(p,U,j),d(p,J,j),d(p,x,j),d(p,F,j)},p:z,d(p){p&&(a(e),a(n),a(o),a(t),a(g),a(U),a(J),a(x),a(F))}}}function Uo(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function xo(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, TFConvBertForSequenceClassification
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> tensorflow <span class="hljs-keyword">as</span> tf

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = TFConvBertForSequenceClassification.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(<span class="hljs-string">"Hello, my dog is cute"</span>, return_tensors=<span class="hljs-string">"tf"</span>)

	<span class="hljs-meta">>>> </span>logits = model(**inputs).logits

	<span class="hljs-meta">>>> </span>predicted_class_id = <span class="hljs-built_in">int</span>(tf.math.argmax(logits, axis=-<span class="hljs-number">1</span>)[<span class="hljs-number">0</span>])`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function zo(C){let e,f;return e=new je({props:{code:"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",highlighted:'<span class="hljs-meta">>>> </span><span class="hljs-comment"># To train a model on `num_labels` classes, you can pass `num_labels=num_labels` to `.from_pretrained(...)`</span>\n<span class="hljs-meta">>>> </span>num_labels = <span class="hljs-built_in">len</span>(model.config.id2label)\n<span class="hljs-meta">>>> </span>model = TFConvBertForSequenceClassification.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>, num_labels=num_labels)\n\n<span class="hljs-meta">>>> </span>labels = tf.constant(<span class="hljs-number">1</span>)\n<span class="hljs-meta">>>> </span>loss = model(**inputs, labels=labels).loss',wrap:!1}}),{c(){T(e.$$.fragment)},l(n){v(e.$$.fragment,n)},m(n,o){M(e,n,o),f=!0},p:z,i(n){f\|\|(y(e.$$.fragment,n),f=!0)},o(n){k(e.$$.fragment,n),f=!1},d(n){w(e,n)}}}function Io(C){let e,f="TensorFlow models and layers in <code>transformers</code> accept two formats as input:",n,o,b="<li>having all inputs as keyword arguments (like PyTorch models), or</li> <li>having all inputs as a list, tuple or dict in the first positional argument.</li>",t,g,E=`The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
	and layers. Because of this support, when using methods like <code>model.fit()</code> things should “just work” for you - just
	pass your inputs and labels in any format that <code>model.fit()</code> supports! If, however, you want to use the second
	format outside of Keras methods like <code>fit()</code> and <code>predict()</code>, such as when creating your own layers or models with
	the Keras <code>Functional</code> API, there are three possibilities you can use to gather all the input Tensors in the first
	positional argument:`,U,J,R=`<li>a single Tensor with <code>input_ids</code> only and nothing else: <code>model(input_ids)</code></li> <li>a list of varying length with one or several input Tensors IN THE ORDER given in the docstring:
	<code>model([input_ids, attention_mask])</code> or <code>model([input_ids, attention_mask, token_type_ids])</code></li> <li>a dictionary with one or several input Tensors associated to the input names given in the docstring:
	<code>model({"input_ids": input_ids, "token_type_ids": token_type_ids})</code></li>`,x,F,V=`Note that when creating models and layers with
	<a href="https://keras.io/guides/making_new_layers_and_models_via_subclassing/" rel="nofollow">subclassing</a> then you don’t need to worry
	about any of this, as you can just pass inputs like you would to any other Python function!`;return{c(){e=m("p"),e.innerHTML=f,n=r(),o=m("ul"),o.innerHTML=b,t=r(),g=m("p"),g.innerHTML=E,U=r(),J=m("ul"),J.innerHTML=R,x=r(),F=m("p"),F.innerHTML=V},l(p){e=u(p,"P",{"data-svelte-h":!0}),_(e)!=="svelte-1ajbfxg"&&(e.innerHTML=f),n=i(p),o=u(p,"UL",{"data-svelte-h":!0}),_(o)!=="svelte-qm1t26"&&(o.innerHTML=b),t=i(p),g=u(p,"P",{"data-svelte-h":!0}),_(g)!=="svelte-1v9qsc5"&&(g.innerHTML=E),U=i(p),J=u(p,"UL",{"data-svelte-h":!0}),_(J)!=="svelte-15scerc"&&(J.innerHTML=R),x=i(p),F=u(p,"P",{"data-svelte-h":!0}),_(F)!=="svelte-1an3odd"&&(F.innerHTML=V)},m(p,j){d(p,e,j),d(p,n,j),d(p,o,j),d(p,t,j),d(p,g,j),d(p,U,j),d(p,J,j),d(p,x,j),d(p,F,j)},p:z,d(p){p&&(a(e),a(n),a(o),a(t),a(g),a(U),a(J),a(x),a(F))}}}function Zo(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function Wo(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, TFConvBertForMultipleChoice
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> tensorflow <span class="hljs-keyword">as</span> tf

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = TFConvBertForMultipleChoice.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>prompt = <span class="hljs-string">"In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."</span>
	<span class="hljs-meta">>>> </span>choice0 = <span class="hljs-string">"It is eaten with a fork and a knife."</span>
	<span class="hljs-meta">>>> </span>choice1 = <span class="hljs-string">"It is eaten while held in the hand."</span>

	<span class="hljs-meta">>>> </span>encoding = tokenizer([prompt, prompt], [choice0, choice1], return_tensors=<span class="hljs-string">"tf"</span>, padding=<span class="hljs-literal">True</span>)
	<span class="hljs-meta">>>> </span>inputs = {k: tf.expand_dims(v, <span class="hljs-number">0</span>) <span class="hljs-keyword">for</span> k, v <span class="hljs-keyword">in</span> encoding.items()}
	<span class="hljs-meta">>>> </span>outputs = model(inputs) <span class="hljs-comment"># batch size is 1</span>

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># the linear classifier still needs to be trained</span>
	<span class="hljs-meta">>>> </span>logits = outputs.logits`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function No(C){let e,f="TensorFlow models and layers in <code>transformers</code> accept two formats as input:",n,o,b="<li>having all inputs as keyword arguments (like PyTorch models), or</li> <li>having all inputs as a list, tuple or dict in the first positional argument.</li>",t,g,E=`The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
	and layers. Because of this support, when using methods like <code>model.fit()</code> things should “just work” for you - just
	pass your inputs and labels in any format that <code>model.fit()</code> supports! If, however, you want to use the second
	format outside of Keras methods like <code>fit()</code> and <code>predict()</code>, such as when creating your own layers or models with
	the Keras <code>Functional</code> API, there are three possibilities you can use to gather all the input Tensors in the first
	positional argument:`,U,J,R=`<li>a single Tensor with <code>input_ids</code> only and nothing else: <code>model(input_ids)</code></li> <li>a list of varying length with one or several input Tensors IN THE ORDER given in the docstring:
	<code>model([input_ids, attention_mask])</code> or <code>model([input_ids, attention_mask, token_type_ids])</code></li> <li>a dictionary with one or several input Tensors associated to the input names given in the docstring:
	<code>model({"input_ids": input_ids, "token_type_ids": token_type_ids})</code></li>`,x,F,V=`Note that when creating models and layers with
	<a href="https://keras.io/guides/making_new_layers_and_models_via_subclassing/" rel="nofollow">subclassing</a> then you don’t need to worry
	about any of this, as you can just pass inputs like you would to any other Python function!`;return{c(){e=m("p"),e.innerHTML=f,n=r(),o=m("ul"),o.innerHTML=b,t=r(),g=m("p"),g.innerHTML=E,U=r(),J=m("ul"),J.innerHTML=R,x=r(),F=m("p"),F.innerHTML=V},l(p){e=u(p,"P",{"data-svelte-h":!0}),_(e)!=="svelte-1ajbfxg"&&(e.innerHTML=f),n=i(p),o=u(p,"UL",{"data-svelte-h":!0}),_(o)!=="svelte-qm1t26"&&(o.innerHTML=b),t=i(p),g=u(p,"P",{"data-svelte-h":!0}),_(g)!=="svelte-1v9qsc5"&&(g.innerHTML=E),U=i(p),J=u(p,"UL",{"data-svelte-h":!0}),_(J)!=="svelte-15scerc"&&(J.innerHTML=R),x=i(p),F=u(p,"P",{"data-svelte-h":!0}),_(F)!=="svelte-1an3odd"&&(F.innerHTML=V)},m(p,j){d(p,e,j),d(p,n,j),d(p,o,j),d(p,t,j),d(p,g,j),d(p,U,j),d(p,J,j),d(p,x,j),d(p,F,j)},p:z,d(p){p&&(a(e),a(n),a(o),a(t),a(g),a(U),a(J),a(x),a(F))}}}function qo(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function Ro(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, TFConvBertForTokenClassification
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> tensorflow <span class="hljs-keyword">as</span> tf

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = TFConvBertForTokenClassification.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>inputs = tokenizer(
	<span class="hljs-meta">... </span> <span class="hljs-string">"HuggingFace is a company based in Paris and New York"</span>, add_special_tokens=<span class="hljs-literal">False</span>, return_tensors=<span class="hljs-string">"tf"</span>
	<span class="hljs-meta">... </span>)

	<span class="hljs-meta">>>> </span>logits = model(**inputs).logits
	<span class="hljs-meta">>>> </span>predicted_token_class_ids = tf.math.argmax(logits, axis=-<span class="hljs-number">1</span>)

	<span class="hljs-meta">>>> </span><span class="hljs-comment"># Note that tokens are classified rather then input words which means that</span>
	<span class="hljs-meta">>>> </span><span class="hljs-comment"># there might be more predicted token classes than words.</span>
	<span class="hljs-meta">>>> </span><span class="hljs-comment"># Multiple token classes might account for the same word</span>
	<span class="hljs-meta">>>> </span>predicted_tokens_classes = [model.config.id2label[t] <span class="hljs-keyword">for</span> t <span class="hljs-keyword">in</span> predicted_token_class_ids[<span class="hljs-number">0</span>].numpy().tolist()]`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function Vo(C){let e,f;return e=new je({props:{code:"bGFiZWxzJTIwJTNEJTIwcHJlZGljdGVkX3Rva2VuX2NsYXNzX2lkcyUwQWxvc3MlMjAlM0QlMjB0Zi5tYXRoLnJlZHVjZV9tZWFuKG1vZGVsKCoqaW5wdXRzJTJDJTIwbGFiZWxzJTNEbGFiZWxzKS5sb3NzKQ==",highlighted:`<span class="hljs-meta">>>> </span>labels = predicted_token_class_ids
	<span class="hljs-meta">>>> </span>loss = tf.math.reduce_mean(model(**inputs, labels=labels).loss)`,wrap:!1}}),{c(){T(e.$$.fragment)},l(n){v(e.$$.fragment,n)},m(n,o){M(e,n,o),f=!0},p:z,i(n){f\|\|(y(e.$$.fragment,n),f=!0)},o(n){k(e.$$.fragment,n),f=!1},d(n){w(e,n)}}}function Go(C){let e,f="TensorFlow models and layers in <code>transformers</code> accept two formats as input:",n,o,b="<li>having all inputs as keyword arguments (like PyTorch models), or</li> <li>having all inputs as a list, tuple or dict in the first positional argument.</li>",t,g,E=`The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
	and layers. Because of this support, when using methods like <code>model.fit()</code> things should “just work” for you - just
	pass your inputs and labels in any format that <code>model.fit()</code> supports! If, however, you want to use the second
	format outside of Keras methods like <code>fit()</code> and <code>predict()</code>, such as when creating your own layers or models with
	the Keras <code>Functional</code> API, there are three possibilities you can use to gather all the input Tensors in the first
	positional argument:`,U,J,R=`<li>a single Tensor with <code>input_ids</code> only and nothing else: <code>model(input_ids)</code></li> <li>a list of varying length with one or several input Tensors IN THE ORDER given in the docstring:
	<code>model([input_ids, attention_mask])</code> or <code>model([input_ids, attention_mask, token_type_ids])</code></li> <li>a dictionary with one or several input Tensors associated to the input names given in the docstring:
	<code>model({"input_ids": input_ids, "token_type_ids": token_type_ids})</code></li>`,x,F,V=`Note that when creating models and layers with
	<a href="https://keras.io/guides/making_new_layers_and_models_via_subclassing/" rel="nofollow">subclassing</a> then you don’t need to worry
	about any of this, as you can just pass inputs like you would to any other Python function!`;return{c(){e=m("p"),e.innerHTML=f,n=r(),o=m("ul"),o.innerHTML=b,t=r(),g=m("p"),g.innerHTML=E,U=r(),J=m("ul"),J.innerHTML=R,x=r(),F=m("p"),F.innerHTML=V},l(p){e=u(p,"P",{"data-svelte-h":!0}),_(e)!=="svelte-1ajbfxg"&&(e.innerHTML=f),n=i(p),o=u(p,"UL",{"data-svelte-h":!0}),_(o)!=="svelte-qm1t26"&&(o.innerHTML=b),t=i(p),g=u(p,"P",{"data-svelte-h":!0}),_(g)!=="svelte-1v9qsc5"&&(g.innerHTML=E),U=i(p),J=u(p,"UL",{"data-svelte-h":!0}),_(J)!=="svelte-15scerc"&&(J.innerHTML=R),x=i(p),F=u(p,"P",{"data-svelte-h":!0}),_(F)!=="svelte-1an3odd"&&(F.innerHTML=V)},m(p,j){d(p,e,j),d(p,n,j),d(p,o,j),d(p,t,j),d(p,g,j),d(p,U,j),d(p,J,j),d(p,x,j),d(p,F,j)},p:z,d(p){p&&(a(e),a(n),a(o),a(t),a(g),a(U),a(J),a(x),a(F))}}}function Ho(C){let e,f=`Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code>
	instance afterwards instead of this since the former takes care of running the pre and post processing steps while
	the latter silently ignores them.`;return{c(){e=m("p"),e.innerHTML=f},l(n){e=u(n,"P",{"data-svelte-h":!0}),_(e)!=="svelte-fincs2"&&(e.innerHTML=f)},m(n,o){d(n,e,o)},p:z,d(n){n&&a(e)}}}function Lo(C){let e,f="Example:",n,o,b;return o=new je({props:{code:"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",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, TFConvBertForQuestionAnswering
	<span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> tensorflow <span class="hljs-keyword">as</span> tf

	<span class="hljs-meta">>>> </span>tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)
	<span class="hljs-meta">>>> </span>model = TFConvBertForQuestionAnswering.from_pretrained(<span class="hljs-string">"YituTech/conv-bert-base"</span>)

	<span class="hljs-meta">>>> </span>question, text = <span class="hljs-string">"Who was Jim Henson?"</span>, <span class="hljs-string">"Jim Henson was a nice puppet"</span>

	<span class="hljs-meta">>>> </span>inputs = tokenizer(question, text, return_tensors=<span class="hljs-string">"tf"</span>)
	<span class="hljs-meta">>>> </span>outputs = model(**inputs)

	<span class="hljs-meta">>>> </span>answer_start_index = <span class="hljs-built_in">int</span>(tf.math.argmax(outputs.start_logits, axis=-<span class="hljs-number">1</span>)[<span class="hljs-number">0</span>])
	<span class="hljs-meta">>>> </span>answer_end_index = <span class="hljs-built_in">int</span>(tf.math.argmax(outputs.end_logits, axis=-<span class="hljs-number">1</span>)[<span class="hljs-number">0</span>])

	<span class="hljs-meta">>>> </span>predict_answer_tokens = inputs.input_ids[<span class="hljs-number">0</span>, answer_start_index : answer_end_index + <span class="hljs-number">1</span>]`,wrap:!1}}),{c(){e=m("p"),e.textContent=f,n=r(),T(o.$$.fragment)},l(t){e=u(t,"P",{"data-svelte-h":!0}),_(e)!=="svelte-11lpom8"&&(e.textContent=f),n=i(t),v(o.$$.fragment,t)},m(t,g){d(t,e,g),d(t,n,g),M(o,t,g),b=!0},p:z,i(t){b\|\|(y(o.$$.fragment,t),b=!0)},o(t){k(o.$$.fragment,t),b=!1},d(t){t&&(a(e),a(n)),w(o,t)}}}function Xo(C){let e,f;return e=new je({props:{code:"JTIzJTIwdGFyZ2V0JTIwaXMlMjAlMjJuaWNlJTIwcHVwcGV0JTIyJTBBdGFyZ2V0X3N0YXJ0X2luZGV4JTIwJTNEJTIwdGYuY29uc3RhbnQoJTVCMTQlNUQpJTBBdGFyZ2V0X2VuZF9pbmRleCUyMCUzRCUyMHRmLmNvbnN0YW50KCU1QjE1JTVEKSUwQSUwQW91dHB1dHMlMjAlM0QlMjBtb2RlbCgqKmlucHV0cyUyQyUyMHN0YXJ0X3Bvc2l0aW9ucyUzRHRhcmdldF9zdGFydF9pbmRleCUyQyUyMGVuZF9wb3NpdGlvbnMlM0R0YXJnZXRfZW5kX2luZGV4KSUwQWxvc3MlMjAlM0QlMjB0Zi5tYXRoLnJlZHVjZV9tZWFuKG91dHB1dHMubG9zcyk=",highlighted:`<span class="hljs-meta">>>> </span><span class="hljs-comment"># target is "nice puppet"</span>
	<span class="hljs-meta">>>> </span>target_start_index = tf.constant([<span class="hljs-number">14</span>])
	<span class="hljs-meta">>>> </span>target_end_index = tf.constant([<span class="hljs-number">15</span>])

	<span class="hljs-meta">>>> </span>outputs = model(**inputs, start_positions=target_start_index, end_positions=target_end_index)
	<span class="hljs-meta">>>> </span>loss = tf.math.reduce_mean(outputs.loss)`,wrap:!1}}),{c(){T(e.$$.fragment)},l(n){v(e.$$.fragment,n)},m(n,o){M(e,n,o),f=!0},p:z,i(n){f\|\|(y(e.$$.fragment,n),f=!0)},o(n){k(e.$$.fragment,n),f=!1},d(n){w(e,n)}}}function Eo(C){let e,f,n,o,b,t,g="The bare ConvBERT Model transformer outputting raw hidden-states without any specific head on top.",E,U,J=`This model inherits from <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.TFPreTrainedModel">TFPreTrainedModel</a>. Check the superclass documentation for the generic methods the
	library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
	etc.)`,R,x,F=`This model is also a <a href="https://www.tensorflow.org/api_docs/python/tf/keras/Model" rel="nofollow">keras.Model</a> subclass. Use it
	as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and
	behavior.`,V,p,j,P,Te,jt,et,fe='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertModel">TFConvBertModel</a> forward method, overrides the <code>__call__</code> special method.',Jt,Ye,qe,tt,nt,le,Se,H,Je,Ft,Et,ot="ConvBERT Model with a <code>language modeling</code> head on top.",ct,ve,_e=`This model inherits from <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.TFPreTrainedModel">TFPreTrainedModel</a>. Check the superclass documentation for the generic methods the
	library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
	etc.)`,Fe,pt,de=`This model is also a <a href="https://www.tensorflow.org/api_docs/python/tf/keras/Model" rel="nofollow">keras.Model</a> subclass. Use it
	as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and
	behavior.`,Ut,st,Ae,Me,mt,Ue,xt,zt='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertForMaskedLM">TFConvBertForMaskedLM</a> forward method, overrides the <code>__call__</code> special method.',ce,I,ut,xe,Qt,ht,Pe,Re,at,K,S,rt,Ve,It="ConvBERT Model transformer with a sequence classification/regression head on top e.g., for GLUE tasks.",ye,Ge,Yt=`This model inherits from <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.TFPreTrainedModel">TFPreTrainedModel</a>. Check the superclass documentation for the generic methods the
	library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
	etc.)`,St,At,ft=`This model is also a <a href="https://www.tensorflow.org/api_docs/python/tf/keras/Model" rel="nofollow">keras.Model</a> subclass. Use it
	as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and
	behavior.`,He,gt,Zt,G,ze,ke,be,pn='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertForSequenceClassification">TFConvBertForSequenceClassification</a> forward method, overrides the <code>__call__</code> special method.',Wt,we,sn,D,nn,pe,on,ee,Ie,L,Ze,Nt,_t,te=`ConvBERT Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a
	softmax) e.g. for RocStories/SWAG tasks.`,qt,Rt,ge=`This model inherits from <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.TFPreTrainedModel">TFPreTrainedModel</a>. Check the superclass documentation for the generic methods the
	library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
	etc.)`,Vt,Gt,it=`This model is also a <a href="https://www.tensorflow.org/api_docs/python/tf/keras/Model" rel="nofollow">keras.Model</a> subclass. Use it
	as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and
	behavior.`,mn,lt,me,oe,Ce,bt,De,an='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertForMultipleChoice">TFConvBertForMultipleChoice</a> forward method, overrides the <code>__call__</code> special method.',se,We,rn,Q,Tt,vt,Le,ue,Mt,ne,Xe,Pt=`ConvBERT Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
	Named-Entity-Recognition (NER) tasks.`,Dt,Ht,yt=`This model inherits from <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.TFPreTrainedModel">TFPreTrainedModel</a>. Check the superclass documentation for the generic methods the
	library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
	etc.)`,Oe,Ee,dt=`This model is also a <a href="https://www.tensorflow.org/api_docs/python/tf/keras/Model" rel="nofollow">keras.Model</a> subclass. Use it
	as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and
	behavior.`,Lt,h,s,$,X,N,Y,ae='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertForTokenClassification">TFConvBertForTokenClassification</a> forward method, overrides the <code>__call__</code> special method.',re,he,O,A,ie,$e,_n,cn,wn,Ke,bn,$n,vn,Zn=`ConvBERT Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear
	layer on top of the hidden-states output to compute <code>span start logits</code> and <code>span end logits</code>).`,Bn,Mn,Wn=`This model inherits from <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.TFPreTrainedModel">TFPreTrainedModel</a>. Check the superclass documentation for the generic methods the
	library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
	etc.)`,jn,yn,Nn=`This model is also a <a href="https://www.tensorflow.org/api_docs/python/tf/keras/Model" rel="nofollow">keras.Model</a> subclass. Use it
	as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and
	behavior.`,Jn,un,Fn,Xt,Tn,Un,kn,qn='The <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertForQuestionAnswering">TFConvBertForQuestionAnswering</a> forward method, overrides the <code>__call__</code> special method.',xn,hn,zn,fn,In,gn,Cn;return e=new Ne({props:{title:"TFConvBertModel",local:"transformers.TFConvBertModel",headingTag:"h2"}}),o=new q({props:{name:"class transformers.TFConvBertModel",anchor:"transformers.TFConvBertModel",parameters:[{name:"config",val:""},{name:"inputs",val:""},{name:"*kwargs",val:""}],parametersDescription:[{anchor:"transformers.TFConvBertModel.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L843"}}),p=new Qe({props:{$$slots:{default:[ko]},$$scope:{ctx:C}}}),Te=new q({props:{name:"call",anchor:"transformers.TFConvBertModel.call",parameters:[{name:"input_ids",val:": TFModelInputType \| None = None"},{name:"attention_mask",val:": Optional[Union[np.array, tf.Tensor]] = None"},{name:"token_type_ids",val:": Optional[Union[np.array, tf.Tensor]] = None"},{name:"position_ids",val:": Optional[Union[np.array, tf.Tensor]] = None"},{name:"head_mask",val:": Optional[Union[np.array, tf.Tensor]] = None"},{name:"inputs_embeds",val:": tf.Tensor \| None = None"},{name:"output_attentions",val:": Optional[bool] = None"},{name:"output_hidden_states",val:": Optional[bool] = None"},{name:"return_dict",val:": Optional[bool] = None"},{name:"training",val:": bool = False"}],parametersDescription:[{anchor:"transformers.TFConvBertModel.call.input_ids",description:`<strong>input_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.TFConvBertModel.call.attention_mask",description:`<strong>attention_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 for tokens that are <strong>not masked</strong>,</li>
	<li>0 for tokens that are <strong>masked</strong>.</li>
	</ul>
	<p><a href="../glossary#attention-mask">What are attention masks?</a>`,name:"attention_mask"},{anchor:"transformers.TFConvBertModel.call.token_type_ids",description:`<strong>token_type_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Segment token indices to indicate first and second portions of the inputs. Indices are selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>0 corresponds to a <em>sentence A</em> token,</li>
	<li>1 corresponds to a <em>sentence B</em> token.</li>
	</ul>
	<p><a href="../glossary#token-type-ids">What are token type IDs?</a>`,name:"token_type_ids"},{anchor:"transformers.TFConvBertModel.call.position_ids",description:`<strong>position_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Indices of positions of each input sequence tokens in the position embeddings. Selected in the range <code>[0, config.max_position_embeddings - 1]</code>.</p>
	<p><a href="../glossary#position-ids">What are position IDs?</a>`,name:"position_ids"},{anchor:"transformers.TFConvBertModel.call.head_mask",description:`<strong>head_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(num_heads,)</code> or <code>(num_layers, num_heads)</code>, <em>optional</em>) —
	Mask to nullify selected heads of the self-attention modules. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 indicates the head is <strong>not masked</strong>,</li>
	<li>0 indicates the head is <strong>masked</strong>.</li>
	</ul>`,name:"head_mask"},{anchor:"transformers.TFConvBertModel.call.inputs_embeds",description:`<strong>inputs_embeds</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>, <em>optional</em>) —
	Optionally, instead of passing <code>input_ids</code> you can choose to directly pass an embedded representation. This
	is useful if you want more control over how to convert <code>input_ids</code> indices into associated vectors than the
	model’s internal embedding lookup matrix.`,name:"inputs_embeds"},{anchor:"transformers.TFConvBertModel.call.output_attentions",description:`<strong>output_attentions</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the attentions tensors of all attention layers. See <code>attentions</code> under returned
	tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
	config will be used instead.`,name:"output_attentions"},{anchor:"transformers.TFConvBertModel.call.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
	more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
	used instead.`,name:"output_hidden_states"},{anchor:"transformers.TFConvBertModel.call.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return a <a href="/docs/transformers/pr_36839/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple. This argument can be used in
	eager mode, in graph mode the value will always be set to True.`,name:"return_dict"},{anchor:"transformers.TFConvBertModel.call.training",description:`<strong>training</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) —
	Whether or not to use the model in training mode (some modules like dropout modules have different
	behaviors between training and evaluation).`,name:"training"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L853",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFBaseModelOutput"
	>transformers.modeling_tf_outputs.TFBaseModelOutput</a> or a tuple of <code>tf.Tensor</code> (if
	<code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various elements depending on the
	configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>last_hidden_state</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>) — Sequence of hidden-states at the output of the last layer of the model.</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(tf.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>tf.Tensor</code> (one for the output of the embeddings + one for the output of each layer) of shape
	<code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>tf.Tensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFBaseModelOutput"
	>transformers.modeling_tf_outputs.TFBaseModelOutput</a> or <code>tuple(tf.Tensor)</code></p>
	`}}),Ye=new Qe({props:{$$slots:{default:[wo]},$$scope:{ctx:C}}}),tt=new Be({props:{anchor:"transformers.TFConvBertModel.call.example",$$slots:{default:[Co]},$$scope:{ctx:C}}}),le=new Ne({props:{title:"TFConvBertForMaskedLM",local:"transformers.TFConvBertForMaskedLM",headingTag:"h2"}}),Je=new q({props:{name:"class transformers.TFConvBertForMaskedLM",anchor:"transformers.TFConvBertForMaskedLM",parameters:[{name:"config",val:""},{name:"inputs",val:""},{name:"*kwargs",val:""}],parametersDescription:[{anchor:"transformers.TFConvBertForMaskedLM.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L961"}}),st=new Qe({props:{$$slots:{default:[$o]},$$scope:{ctx:C}}}),mt=new q({props:{name:"call",anchor:"transformers.TFConvBertForMaskedLM.call",parameters:[{name:"input_ids",val:": TFModelInputType \| None = None"},{name:"attention_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"token_type_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"position_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"head_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"inputs_embeds",val:": tf.Tensor \| None = None"},{name:"output_attentions",val:": Optional[bool] = None"},{name:"output_hidden_states",val:": Optional[bool] = None"},{name:"return_dict",val:": Optional[bool] = None"},{name:"labels",val:": tf.Tensor \| None = None"},{name:"training",val:": Optional[bool] = False"}],parametersDescription:[{anchor:"transformers.TFConvBertForMaskedLM.call.input_ids",description:`<strong>input_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.TFConvBertForMaskedLM.call.attention_mask",description:`<strong>attention_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 for tokens that are <strong>not masked</strong>,</li>
	<li>0 for tokens that are <strong>masked</strong>.</li>
	</ul>
	<p><a href="../glossary#attention-mask">What are attention masks?</a>`,name:"attention_mask"},{anchor:"transformers.TFConvBertForMaskedLM.call.token_type_ids",description:`<strong>token_type_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Segment token indices to indicate first and second portions of the inputs. Indices are selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>0 corresponds to a <em>sentence A</em> token,</li>
	<li>1 corresponds to a <em>sentence B</em> token.</li>
	</ul>
	<p><a href="../glossary#token-type-ids">What are token type IDs?</a>`,name:"token_type_ids"},{anchor:"transformers.TFConvBertForMaskedLM.call.position_ids",description:`<strong>position_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Indices of positions of each input sequence tokens in the position embeddings. Selected in the range <code>[0, config.max_position_embeddings - 1]</code>.</p>
	<p><a href="../glossary#position-ids">What are position IDs?</a>`,name:"position_ids"},{anchor:"transformers.TFConvBertForMaskedLM.call.head_mask",description:`<strong>head_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(num_heads,)</code> or <code>(num_layers, num_heads)</code>, <em>optional</em>) —
	Mask to nullify selected heads of the self-attention modules. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 indicates the head is <strong>not masked</strong>,</li>
	<li>0 indicates the head is <strong>masked</strong>.</li>
	</ul>`,name:"head_mask"},{anchor:"transformers.TFConvBertForMaskedLM.call.inputs_embeds",description:`<strong>inputs_embeds</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>, <em>optional</em>) —
	Optionally, instead of passing <code>input_ids</code> you can choose to directly pass an embedded representation. This
	is useful if you want more control over how to convert <code>input_ids</code> indices into associated vectors than the
	model’s internal embedding lookup matrix.`,name:"inputs_embeds"},{anchor:"transformers.TFConvBertForMaskedLM.call.output_attentions",description:`<strong>output_attentions</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the attentions tensors of all attention layers. See <code>attentions</code> under returned
	tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
	config will be used instead.`,name:"output_attentions"},{anchor:"transformers.TFConvBertForMaskedLM.call.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
	more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
	used instead.`,name:"output_hidden_states"},{anchor:"transformers.TFConvBertForMaskedLM.call.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return a <a href="/docs/transformers/pr_36839/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple. This argument can be used in
	eager mode, in graph mode the value will always be set to True.`,name:"return_dict"},{anchor:"transformers.TFConvBertForMaskedLM.call.training",description:`<strong>training</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) —
	Whether or not to use the model in training mode (some modules like dropout modules have different
	behaviors between training and evaluation).`,name:"training"},{anchor:"transformers.TFConvBertForMaskedLM.call.labels",description:`<strong>labels</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Labels for computing the masked language modeling loss. Indices should be in <code>[-100, 0, ..., config.vocab_size]</code> (see <code>input_ids</code> docstring) Tokens with indices set to <code>-100</code> are ignored (masked), the
	loss is only computed for the tokens with labels in <code>[0, ..., config.vocab_size]</code>`,name:"labels"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L983",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFMaskedLMOutput"
	>transformers.modeling_tf_outputs.TFMaskedLMOutput</a> or a tuple of <code>tf.Tensor</code> (if
	<code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various elements depending on the
	configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>tf.Tensor</code> of shape <code>(n,)</code>, <em>optional</em>, where n is the number of non-masked labels, returned when <code>labels</code> is provided) — Masked language modeling (MLM) loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, config.vocab_size)</code>) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>tf.Tensor</code> (one for the output of the embeddings + one for the output of each layer) of shape
	<code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>tf.Tensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFMaskedLMOutput"
	>transformers.modeling_tf_outputs.TFMaskedLMOutput</a> or <code>tuple(tf.Tensor)</code></p>
	`}}),I=new Qe({props:{$$slots:{default:[Bo]},$$scope:{ctx:C}}}),xe=new Be({props:{anchor:"transformers.TFConvBertForMaskedLM.call.example",$$slots:{default:[jo]},$$scope:{ctx:C}}}),ht=new Be({props:{anchor:"transformers.TFConvBertForMaskedLM.call.example-2",$$slots:{default:[Jo]},$$scope:{ctx:C}}}),Re=new Ne({props:{title:"TFConvBertForSequenceClassification",local:"transformers.TFConvBertForSequenceClassification",headingTag:"h2"}}),S=new q({props:{name:"class transformers.TFConvBertForSequenceClassification",anchor:"transformers.TFConvBertForSequenceClassification",parameters:[{name:"config",val:""},{name:"inputs",val:""},{name:"*kwargs",val:""}],parametersDescription:[{anchor:"transformers.TFConvBertForSequenceClassification.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1095"}}),gt=new Qe({props:{$$slots:{default:[Fo]},$$scope:{ctx:C}}}),ze=new q({props:{name:"call",anchor:"transformers.TFConvBertForSequenceClassification.call",parameters:[{name:"input_ids",val:": TFModelInputType \| None = None"},{name:"attention_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"token_type_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"position_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"head_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"inputs_embeds",val:": tf.Tensor \| None = None"},{name:"output_attentions",val:": Optional[bool] = None"},{name:"output_hidden_states",val:": Optional[bool] = None"},{name:"return_dict",val:": Optional[bool] = None"},{name:"labels",val:": tf.Tensor \| None = None"},{name:"training",val:": Optional[bool] = False"}],parametersDescription:[{anchor:"transformers.TFConvBertForSequenceClassification.call.input_ids",description:`<strong>input_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.TFConvBertForSequenceClassification.call.attention_mask",description:`<strong>attention_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 for tokens that are <strong>not masked</strong>,</li>
	<li>0 for tokens that are <strong>masked</strong>.</li>
	</ul>
	<p><a href="../glossary#attention-mask">What are attention masks?</a>`,name:"attention_mask"},{anchor:"transformers.TFConvBertForSequenceClassification.call.token_type_ids",description:`<strong>token_type_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Segment token indices to indicate first and second portions of the inputs. Indices are selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>0 corresponds to a <em>sentence A</em> token,</li>
	<li>1 corresponds to a <em>sentence B</em> token.</li>
	</ul>
	<p><a href="../glossary#token-type-ids">What are token type IDs?</a>`,name:"token_type_ids"},{anchor:"transformers.TFConvBertForSequenceClassification.call.position_ids",description:`<strong>position_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Indices of positions of each input sequence tokens in the position embeddings. Selected in the range <code>[0, config.max_position_embeddings - 1]</code>.</p>
	<p><a href="../glossary#position-ids">What are position IDs?</a>`,name:"position_ids"},{anchor:"transformers.TFConvBertForSequenceClassification.call.head_mask",description:`<strong>head_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(num_heads,)</code> or <code>(num_layers, num_heads)</code>, <em>optional</em>) —
	Mask to nullify selected heads of the self-attention modules. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 indicates the head is <strong>not masked</strong>,</li>
	<li>0 indicates the head is <strong>masked</strong>.</li>
	</ul>`,name:"head_mask"},{anchor:"transformers.TFConvBertForSequenceClassification.call.inputs_embeds",description:`<strong>inputs_embeds</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>, <em>optional</em>) —
	Optionally, instead of passing <code>input_ids</code> you can choose to directly pass an embedded representation. This
	is useful if you want more control over how to convert <code>input_ids</code> indices into associated vectors than the
	model’s internal embedding lookup matrix.`,name:"inputs_embeds"},{anchor:"transformers.TFConvBertForSequenceClassification.call.output_attentions",description:`<strong>output_attentions</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the attentions tensors of all attention layers. See <code>attentions</code> under returned
	tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
	config will be used instead.`,name:"output_attentions"},{anchor:"transformers.TFConvBertForSequenceClassification.call.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
	more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
	used instead.`,name:"output_hidden_states"},{anchor:"transformers.TFConvBertForSequenceClassification.call.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return a <a href="/docs/transformers/pr_36839/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple. This argument can be used in
	eager mode, in graph mode the value will always be set to True.`,name:"return_dict"},{anchor:"transformers.TFConvBertForSequenceClassification.call.training",description:`<strong>training</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) —
	Whether or not to use the model in training mode (some modules like dropout modules have different
	behaviors between training and evaluation).`,name:"training"},{anchor:"transformers.TFConvBertForSequenceClassification.call.labels",description:`<strong>labels</strong> (<code>tf.Tensor</code> of shape <code>(batch_size,)</code>, <em>optional</em>) —
	Labels for computing the sequence classification/regression loss. Indices should be in <code>[0, ..., config.num_labels - 1]</code>. If <code>config.num_labels == 1</code> a regression loss is computed (Mean-Square loss), If
	<code>config.num_labels > 1</code> a classification loss is computed (Cross-Entropy).`,name:"labels"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1108",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFSequenceClassifierOutput"
	>transformers.modeling_tf_outputs.TFSequenceClassifierOutput</a> or a tuple of <code>tf.Tensor</code> (if
	<code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various elements depending on the
	configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, )</code>, <em>optional</em>, returned when <code>labels</code> is provided) — Classification (or regression if config.num_labels==1) loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, config.num_labels)</code>) — Classification (or regression if config.num_labels==1) scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>tf.Tensor</code> (one for the output of the embeddings + one for the output of each layer) of shape
	<code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>tf.Tensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFSequenceClassifierOutput"
	>transformers.modeling_tf_outputs.TFSequenceClassifierOutput</a> or <code>tuple(tf.Tensor)</code></p>
	`}}),we=new Qe({props:{$$slots:{default:[Uo]},$$scope:{ctx:C}}}),D=new Be({props:{anchor:"transformers.TFConvBertForSequenceClassification.call.example",$$slots:{default:[xo]},$$scope:{ctx:C}}}),pe=new Be({props:{anchor:"transformers.TFConvBertForSequenceClassification.call.example-2",$$slots:{default:[zo]},$$scope:{ctx:C}}}),ee=new Ne({props:{title:"TFConvBertForMultipleChoice",local:"transformers.TFConvBertForMultipleChoice",headingTag:"h2"}}),Ze=new q({props:{name:"class transformers.TFConvBertForMultipleChoice",anchor:"transformers.TFConvBertForMultipleChoice",parameters:[{name:"config",val:""},{name:"inputs",val:""},{name:"*kwargs",val:""}],parametersDescription:[{anchor:"transformers.TFConvBertForMultipleChoice.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1174"}}),lt=new Qe({props:{$$slots:{default:[Io]},$$scope:{ctx:C}}}),Ce=new q({props:{name:"call",anchor:"transformers.TFConvBertForMultipleChoice.call",parameters:[{name:"input_ids",val:": TFModelInputType \| None = None"},{name:"attention_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"token_type_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"position_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"head_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"inputs_embeds",val:": tf.Tensor \| None = None"},{name:"output_attentions",val:": Optional[bool] = None"},{name:"output_hidden_states",val:": Optional[bool] = None"},{name:"return_dict",val:": Optional[bool] = None"},{name:"labels",val:": tf.Tensor \| None = None"},{name:"training",val:": Optional[bool] = False"}],parametersDescription:[{anchor:"transformers.TFConvBertForMultipleChoice.call.input_ids",description:`<strong>input_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, num_choices, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.TFConvBertForMultipleChoice.call.attention_mask",description:`<strong>attention_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, num_choices, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 for tokens that are <strong>not masked</strong>,</li>
	<li>0 for tokens that are <strong>masked</strong>.</li>
	</ul>
	<p><a href="../glossary#attention-mask">What are attention masks?</a>`,name:"attention_mask"},{anchor:"transformers.TFConvBertForMultipleChoice.call.token_type_ids",description:`<strong>token_type_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, num_choices, sequence_length)</code>, <em>optional</em>) —
	Segment token indices to indicate first and second portions of the inputs. Indices are selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>0 corresponds to a <em>sentence A</em> token,</li>
	<li>1 corresponds to a <em>sentence B</em> token.</li>
	</ul>
	<p><a href="../glossary#token-type-ids">What are token type IDs?</a>`,name:"token_type_ids"},{anchor:"transformers.TFConvBertForMultipleChoice.call.position_ids",description:`<strong>position_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, num_choices, sequence_length)</code>, <em>optional</em>) —
	Indices of positions of each input sequence tokens in the position embeddings. Selected in the range <code>[0, config.max_position_embeddings - 1]</code>.</p>
	<p><a href="../glossary#position-ids">What are position IDs?</a>`,name:"position_ids"},{anchor:"transformers.TFConvBertForMultipleChoice.call.head_mask",description:`<strong>head_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(num_heads,)</code> or <code>(num_layers, num_heads)</code>, <em>optional</em>) —
	Mask to nullify selected heads of the self-attention modules. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 indicates the head is <strong>not masked</strong>,</li>
	<li>0 indicates the head is <strong>masked</strong>.</li>
	</ul>`,name:"head_mask"},{anchor:"transformers.TFConvBertForMultipleChoice.call.inputs_embeds",description:`<strong>inputs_embeds</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, num_choices, sequence_length, hidden_size)</code>, <em>optional</em>) —
	Optionally, instead of passing <code>input_ids</code> you can choose to directly pass an embedded representation. This
	is useful if you want more control over how to convert <code>input_ids</code> indices into associated vectors than the
	model’s internal embedding lookup matrix.`,name:"inputs_embeds"},{anchor:"transformers.TFConvBertForMultipleChoice.call.output_attentions",description:`<strong>output_attentions</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the attentions tensors of all attention layers. See <code>attentions</code> under returned
	tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
	config will be used instead.`,name:"output_attentions"},{anchor:"transformers.TFConvBertForMultipleChoice.call.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
	more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
	used instead.`,name:"output_hidden_states"},{anchor:"transformers.TFConvBertForMultipleChoice.call.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return a <a href="/docs/transformers/pr_36839/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple. This argument can be used in
	eager mode, in graph mode the value will always be set to True.`,name:"return_dict"},{anchor:"transformers.TFConvBertForMultipleChoice.call.training",description:`<strong>training</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) —
	Whether or not to use the model in training mode (some modules like dropout modules have different
	behaviors between training and evaluation).`,name:"training"},{anchor:"transformers.TFConvBertForMultipleChoice.call.labels",description:`<strong>labels</strong> (<code>tf.Tensor</code> of shape <code>(batch_size,)</code>, <em>optional</em>) —
	Labels for computing the multiple choice classification loss. Indices should be in <code>[0, ..., num_choices]</code>
	where <code>num_choices</code> is the size of the second dimension of the input tensors. (See <code>input_ids</code> above)`,name:"labels"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1194",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFMultipleChoiceModelOutput"
	>transformers.modeling_tf_outputs.TFMultipleChoiceModelOutput</a> or a tuple of <code>tf.Tensor</code> (if
	<code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various elements depending on the
	configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>tf.Tensor</code> of shape <em>(batch_size, )</em>, <em>optional</em>, returned when <code>labels</code> is provided) — Classification loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, num_choices)</code>) — <em>num_choices</em> is the second dimension of the input tensors. (see <em>input_ids</em> above).</p>
	<p>Classification scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>tf.Tensor</code> (one for the output of the embeddings + one for the output of each layer) of shape
	<code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>tf.Tensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFMultipleChoiceModelOutput"
	>transformers.modeling_tf_outputs.TFMultipleChoiceModelOutput</a> or <code>tuple(tf.Tensor)</code></p>
	`}}),We=new Qe({props:{$$slots:{default:[Zo]},$$scope:{ctx:C}}}),Q=new Be({props:{anchor:"transformers.TFConvBertForMultipleChoice.call.example",$$slots:{default:[Wo]},$$scope:{ctx:C}}}),vt=new Ne({props:{title:"TFConvBertForTokenClassification",local:"transformers.TFConvBertForTokenClassification",headingTag:"h2"}}),Mt=new q({props:{name:"class transformers.TFConvBertForTokenClassification",anchor:"transformers.TFConvBertForTokenClassification",parameters:[{name:"config",val:""},{name:"inputs",val:""},{name:"*kwargs",val:""}],parametersDescription:[{anchor:"transformers.TFConvBertForTokenClassification.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1282"}}),h=new Qe({props:{$$slots:{default:[No]},$$scope:{ctx:C}}}),X=new q({props:{name:"call",anchor:"transformers.TFConvBertForTokenClassification.call",parameters:[{name:"input_ids",val:": TFModelInputType \| None = None"},{name:"attention_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"token_type_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"position_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"head_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"inputs_embeds",val:": tf.Tensor \| None = None"},{name:"output_attentions",val:": Optional[bool] = None"},{name:"output_hidden_states",val:": Optional[bool] = None"},{name:"return_dict",val:": Optional[bool] = None"},{name:"labels",val:": tf.Tensor \| None = None"},{name:"training",val:": Optional[bool] = False"}],parametersDescription:[{anchor:"transformers.TFConvBertForTokenClassification.call.input_ids",description:`<strong>input_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.TFConvBertForTokenClassification.call.attention_mask",description:`<strong>attention_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 for tokens that are <strong>not masked</strong>,</li>
	<li>0 for tokens that are <strong>masked</strong>.</li>
	</ul>
	<p><a href="../glossary#attention-mask">What are attention masks?</a>`,name:"attention_mask"},{anchor:"transformers.TFConvBertForTokenClassification.call.token_type_ids",description:`<strong>token_type_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Segment token indices to indicate first and second portions of the inputs. Indices are selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>0 corresponds to a <em>sentence A</em> token,</li>
	<li>1 corresponds to a <em>sentence B</em> token.</li>
	</ul>
	<p><a href="../glossary#token-type-ids">What are token type IDs?</a>`,name:"token_type_ids"},{anchor:"transformers.TFConvBertForTokenClassification.call.position_ids",description:`<strong>position_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Indices of positions of each input sequence tokens in the position embeddings. Selected in the range <code>[0, config.max_position_embeddings - 1]</code>.</p>
	<p><a href="../glossary#position-ids">What are position IDs?</a>`,name:"position_ids"},{anchor:"transformers.TFConvBertForTokenClassification.call.head_mask",description:`<strong>head_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(num_heads,)</code> or <code>(num_layers, num_heads)</code>, <em>optional</em>) —
	Mask to nullify selected heads of the self-attention modules. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 indicates the head is <strong>not masked</strong>,</li>
	<li>0 indicates the head is <strong>masked</strong>.</li>
	</ul>`,name:"head_mask"},{anchor:"transformers.TFConvBertForTokenClassification.call.inputs_embeds",description:`<strong>inputs_embeds</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>, <em>optional</em>) —
	Optionally, instead of passing <code>input_ids</code> you can choose to directly pass an embedded representation. This
	is useful if you want more control over how to convert <code>input_ids</code> indices into associated vectors than the
	model’s internal embedding lookup matrix.`,name:"inputs_embeds"},{anchor:"transformers.TFConvBertForTokenClassification.call.output_attentions",description:`<strong>output_attentions</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the attentions tensors of all attention layers. See <code>attentions</code> under returned
	tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
	config will be used instead.`,name:"output_attentions"},{anchor:"transformers.TFConvBertForTokenClassification.call.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
	more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
	used instead.`,name:"output_hidden_states"},{anchor:"transformers.TFConvBertForTokenClassification.call.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return a <a href="/docs/transformers/pr_36839/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple. This argument can be used in
	eager mode, in graph mode the value will always be set to True.`,name:"return_dict"},{anchor:"transformers.TFConvBertForTokenClassification.call.training",description:`<strong>training</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) —
	Whether or not to use the model in training mode (some modules like dropout modules have different
	behaviors between training and evaluation).`,name:"training"},{anchor:"transformers.TFConvBertForTokenClassification.call.labels",description:`<strong>labels</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Labels for computing the token classification loss. Indices should be in <code>[0, ..., config.num_labels - 1]</code>.`,name:"labels"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1304",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFTokenClassifierOutput"
	>transformers.modeling_tf_outputs.TFTokenClassifierOutput</a> or a tuple of <code>tf.Tensor</code> (if
	<code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various elements depending on the
	configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>tf.Tensor</code> of shape <code>(n,)</code>, <em>optional</em>, where n is the number of unmasked labels, returned when <code>labels</code> is provided) — Classification loss.</p>
	</li>
	<li>
	<p><strong>logits</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, config.num_labels)</code>) — Classification scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>tf.Tensor</code> (one for the output of the embeddings + one for the output of each layer) of shape
	<code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>tf.Tensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFTokenClassifierOutput"
	>transformers.modeling_tf_outputs.TFTokenClassifierOutput</a> or <code>tuple(tf.Tensor)</code></p>
	`}}),he=new Qe({props:{$$slots:{default:[qo]},$$scope:{ctx:C}}}),A=new Be({props:{anchor:"transformers.TFConvBertForTokenClassification.call.example",$$slots:{default:[Ro]},$$scope:{ctx:C}}}),$e=new Be({props:{anchor:"transformers.TFConvBertForTokenClassification.call.example-2",$$slots:{default:[Vo]},$$scope:{ctx:C}}}),cn=new Ne({props:{title:"TFConvBertForQuestionAnswering",local:"transformers.TFConvBertForQuestionAnswering",headingTag:"h2"}}),bn=new q({props:{name:"class transformers.TFConvBertForQuestionAnswering",anchor:"transformers.TFConvBertForQuestionAnswering",parameters:[{name:"config",val:""},{name:"inputs",val:""},{name:"*kwargs",val:""}],parametersDescription:[{anchor:"transformers.TFConvBertForQuestionAnswering.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig">ConvBertConfig</a>) — Model configuration class with all the parameters of the model.
	Initializing with a config file does not load the weights associated with the model, only the
	configuration. Check out the <a href="/docs/transformers/pr_36839/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1369"}}),un=new Qe({props:{$$slots:{default:[Go]},$$scope:{ctx:C}}}),Tn=new q({props:{name:"call",anchor:"transformers.TFConvBertForQuestionAnswering.call",parameters:[{name:"input_ids",val:": TFModelInputType \| None = None"},{name:"attention_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"token_type_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"position_ids",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"head_mask",val:": np.ndarray \| tf.Tensor \| None = None"},{name:"inputs_embeds",val:": tf.Tensor \| None = None"},{name:"output_attentions",val:": Optional[bool] = None"},{name:"output_hidden_states",val:": Optional[bool] = None"},{name:"return_dict",val:": Optional[bool] = None"},{name:"start_positions",val:": tf.Tensor \| None = None"},{name:"end_positions",val:": tf.Tensor \| None = None"},{name:"training",val:": Optional[bool] = False"}],parametersDescription:[{anchor:"transformers.TFConvBertForQuestionAnswering.call.input_ids",description:`<strong>input_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>) —
	Indices of input sequence tokens in the vocabulary.</p>
	<p>Indices can be obtained using <a href="/docs/transformers/pr_36839/en/model_doc/auto#transformers.AutoTokenizer">AutoTokenizer</a>. See <a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.__call__">PreTrainedTokenizer.<strong>call</strong>()</a> and
	<a href="/docs/transformers/pr_36839/en/internal/tokenization_utils#transformers.PreTrainedTokenizerBase.encode">PreTrainedTokenizer.encode()</a> for details.</p>
	<p><a href="../glossary#input-ids">What are input IDs?</a>`,name:"input_ids"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.attention_mask",description:`<strong>attention_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Mask to avoid performing attention on padding token indices. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 for tokens that are <strong>not masked</strong>,</li>
	<li>0 for tokens that are <strong>masked</strong>.</li>
	</ul>
	<p><a href="../glossary#attention-mask">What are attention masks?</a>`,name:"attention_mask"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.token_type_ids",description:`<strong>token_type_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Segment token indices to indicate first and second portions of the inputs. Indices are selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>0 corresponds to a <em>sentence A</em> token,</li>
	<li>1 corresponds to a <em>sentence B</em> token.</li>
	</ul>
	<p><a href="../glossary#token-type-ids">What are token type IDs?</a>`,name:"token_type_ids"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.position_ids",description:`<strong>position_ids</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>, <em>optional</em>) —
	Indices of positions of each input sequence tokens in the position embeddings. Selected in the range <code>[0, config.max_position_embeddings - 1]</code>.</p>
	<p><a href="../glossary#position-ids">What are position IDs?</a>`,name:"position_ids"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.head_mask",description:`<strong>head_mask</strong> (<code>Numpy array</code> or <code>tf.Tensor</code> of shape <code>(num_heads,)</code> or <code>(num_layers, num_heads)</code>, <em>optional</em>) —
	Mask to nullify selected heads of the self-attention modules. Mask values selected in <code>[0, 1]</code>:</p>
	<ul>
	<li>1 indicates the head is <strong>not masked</strong>,</li>
	<li>0 indicates the head is <strong>masked</strong>.</li>
	</ul>`,name:"head_mask"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.inputs_embeds",description:`<strong>inputs_embeds</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>, <em>optional</em>) —
	Optionally, instead of passing <code>input_ids</code> you can choose to directly pass an embedded representation. This
	is useful if you want more control over how to convert <code>input_ids</code> indices into associated vectors than the
	model’s internal embedding lookup matrix.`,name:"inputs_embeds"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.output_attentions",description:`<strong>output_attentions</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the attentions tensors of all attention layers. See <code>attentions</code> under returned
	tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
	config will be used instead.`,name:"output_attentions"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
	more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
	used instead.`,name:"output_hidden_states"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to return a <a href="/docs/transformers/pr_36839/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple. This argument can be used in
	eager mode, in graph mode the value will always be set to True.`,name:"return_dict"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.training",description:`<strong>training</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) —
	Whether or not to use the model in training mode (some modules like dropout modules have different
	behaviors between training and evaluation).`,name:"training"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.start_positions",description:`<strong>start_positions</strong> (<code>tf.Tensor</code> of shape <code>(batch_size,)</code>, <em>optional</em>) —
	Labels for position (index) of the start of the labelled span for computing the token classification loss.
	Positions are clamped to the length of the sequence (<code>sequence_length</code>). Position outside of the sequence
	are not taken into account for computing the loss.`,name:"start_positions"},{anchor:"transformers.TFConvBertForQuestionAnswering.call.end_positions",description:`<strong>end_positions</strong> (<code>tf.Tensor</code> of shape <code>(batch_size,)</code>, <em>optional</em>) —
	Labels for position (index) of the end of the labelled span for computing the token classification loss.
	Positions are clamped to the length of the sequence (<code>sequence_length</code>). Position outside of the sequence
	are not taken into account for computing the loss.`,name:"end_positions"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/modeling_tf_convbert.py#L1387",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A <a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFQuestionAnsweringModelOutput"
	>transformers.modeling_tf_outputs.TFQuestionAnsweringModelOutput</a> or a tuple of <code>tf.Tensor</code> (if
	<code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various elements depending on the
	configuration (<a
	href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertConfig"
	>ConvBertConfig</a>) and inputs.</p>
	<ul>
	<li>
	<p><strong>loss</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, )</code>, <em>optional</em>, returned when <code>start_positions</code> and <code>end_positions</code> are provided) — Total span extraction loss is the sum of a Cross-Entropy for the start and end positions.</p>
	</li>
	<li>
	<p><strong>start_logits</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>) — Span-start scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>end_logits</strong> (<code>tf.Tensor</code> of shape <code>(batch_size, sequence_length)</code>) — Span-end scores (before SoftMax).</p>
	</li>
	<li>
	<p><strong>hidden_states</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>tf.Tensor</code> (one for the output of the embeddings + one for the output of each layer) of shape
	<code>(batch_size, sequence_length, hidden_size)</code>.</p>
	<p>Hidden-states of the model at the output of each layer plus the initial embedding outputs.</p>
	</li>
	<li>
	<p><strong>attentions</strong> (<code>tuple(tf.Tensor)</code>, <em>optional</em>, returned when <code>output_attentions=True</code> is passed or when <code>config.output_attentions=True</code>) — Tuple of <code>tf.Tensor</code> (one for each layer) of shape <code>(batch_size, num_heads, sequence_length, sequence_length)</code>.</p>
	<p>Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
	heads.</p>
	</li>
	</ul>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><a
	href="/docs/transformers/pr_36839/en/main_classes/output#transformers.modeling_tf_outputs.TFQuestionAnsweringModelOutput"
	>transformers.modeling_tf_outputs.TFQuestionAnsweringModelOutput</a> or <code>tuple(tf.Tensor)</code></p>
	`}}),hn=new Qe({props:{$$slots:{default:[Ho]},$$scope:{ctx:C}}}),fn=new Be({props:{anchor:"transformers.TFConvBertForQuestionAnswering.call.example",$$slots:{default:[Lo]},$$scope:{ctx:C}}}),gn=new Be({props:{anchor:"transformers.TFConvBertForQuestionAnswering.call.example-2",$$slots:{default:[Xo]},$$scope:{ctx:C}}}),{c(){T(e.$$.fragment),f=r(),n=m("div"),T(o.$$.fragment),b=r(),t=m("p"),t.textContent=g,E=r(),U=m("p"),U.innerHTML=J,R=r(),x=m("p"),x.innerHTML=F,V=r(),T(p.$$.fragment),j=r(),P=m("div"),T(Te.$$.fragment),jt=r(),et=m("p"),et.innerHTML=fe,Jt=r(),T(Ye.$$.fragment),qe=r(),T(tt.$$.fragment),nt=r(),T(le.$$.fragment),Se=r(),H=m("div"),T(Je.$$.fragment),Ft=r(),Et=m("p"),Et.innerHTML=ot,ct=r(),ve=m("p"),ve.innerHTML=_e,Fe=r(),pt=m("p"),pt.innerHTML=de,Ut=r(),T(st.$$.fragment),Ae=r(),Me=m("div"),T(mt.$$.fragment),Ue=r(),xt=m("p"),xt.innerHTML=zt,ce=r(),T(I.$$.fragment),ut=r(),T(xe.$$.fragment),Qt=r(),T(ht.$$.fragment),Pe=r(),T(Re.$$.fragment),at=r(),K=m("div"),T(S.$$.fragment),rt=r(),Ve=m("p"),Ve.textContent=It,ye=r(),Ge=m("p"),Ge.innerHTML=Yt,St=r(),At=m("p"),At.innerHTML=ft,He=r(),T(gt.$$.fragment),Zt=r(),G=m("div"),T(ze.$$.fragment),ke=r(),be=m("p"),be.innerHTML=pn,Wt=r(),T(we.$$.fragment),sn=r(),T(D.$$.fragment),nn=r(),T(pe.$$.fragment),on=r(),T(ee.$$.fragment),Ie=r(),L=m("div"),T(Ze.$$.fragment),Nt=r(),_t=m("p"),_t.textContent=te,qt=r(),Rt=m("p"),Rt.innerHTML=ge,Vt=r(),Gt=m("p"),Gt.innerHTML=it,mn=r(),T(lt.$$.fragment),me=r(),oe=m("div"),T(Ce.$$.fragment),bt=r(),De=m("p"),De.innerHTML=an,se=r(),T(We.$$.fragment),rn=r(),T(Q.$$.fragment),Tt=r(),T(vt.$$.fragment),Le=r(),ue=m("div"),T(Mt.$$.fragment),ne=r(),Xe=m("p"),Xe.textContent=Pt,Dt=r(),Ht=m("p"),Ht.innerHTML=yt,Oe=r(),Ee=m("p"),Ee.innerHTML=dt,Lt=r(),T(h.$$.fragment),s=r(),$=m("div"),T(X.$$.fragment),N=r(),Y=m("p"),Y.innerHTML=ae,re=r(),T(he.$$.fragment),O=r(),T(A.$$.fragment),ie=r(),T($e.$$.fragment),_n=r(),T(cn.$$.fragment),wn=r(),Ke=m("div"),T(bn.$$.fragment),$n=r(),vn=m("p"),vn.innerHTML=Zn,Bn=r(),Mn=m("p"),Mn.innerHTML=Wn,jn=r(),yn=m("p"),yn.innerHTML=Nn,Jn=r(),T(un.$$.fragment),Fn=r(),Xt=m("div"),T(Tn.$$.fragment),Un=r(),kn=m("p"),kn.innerHTML=qn,xn=r(),T(hn.$$.fragment),zn=r(),T(fn.$$.fragment),In=r(),T(gn.$$.fragment),this.h()},l(c){v(e.$$.fragment,c),f=i(c),n=u(c,"DIV",{class:!0});var B=W(n);v(o.$$.fragment,B),b=i(B),t=u(B,"P",{"data-svelte-h":!0}),_(t)!=="svelte-wivmi6"&&(t.textContent=g),E=i(B),U=u(B,"P",{"data-svelte-h":!0}),_(U)!=="svelte-1pln4c3"&&(U.innerHTML=J),R=i(B),x=u(B,"P",{"data-svelte-h":!0}),_(x)!=="svelte-1be7e3c"&&(x.innerHTML=F),V=i(B),v(p.$$.fragment,B),j=i(B),P=u(B,"DIV",{class:!0});var ln=W(P);v(Te.$$.fragment,ln),jt=i(ln),et=u(ln,"P",{"data-svelte-h":!0}),_(et)!=="svelte-1a8wz21"&&(et.innerHTML=fe),Jt=i(ln),v(Ye.$$.fragment,ln),qe=i(ln),v(tt.$$.fragment,ln),ln.forEach(a),B.forEach(a),nt=i(c),v(le.$$.fragment,c),Se=i(c),H=u(c,"DIV",{class:!0});var kt=W(H);v(Je.$$.fragment,kt),Ft=i(kt),Et=u(kt,"P",{"data-svelte-h":!0}),_(Et)!=="svelte-oaylkh"&&(Et.innerHTML=ot),ct=i(kt),ve=u(kt,"P",{"data-svelte-h":!0}),_(ve)!=="svelte-1pln4c3"&&(ve.innerHTML=_e),Fe=i(kt),pt=u(kt,"P",{"data-svelte-h":!0}),_(pt)!=="svelte-1be7e3c"&&(pt.innerHTML=de),Ut=i(kt),v(st.$$.fragment,kt),Ae=i(kt),Me=u(kt,"DIV",{class:!0});var Ot=W(Me);v(mt.$$.fragment,Ot),Ue=i(Ot),xt=u(Ot,"P",{"data-svelte-h":!0}),_(xt)!=="svelte-1yfc3d"&&(xt.innerHTML=zt),ce=i(Ot),v(I.$$.fragment,Ot),ut=i(Ot),v(xe.$$.fragment,Ot),Qt=i(Ot),v(ht.$$.fragment,Ot),Ot.forEach(a),kt.forEach(a),Pe=i(c),v(Re.$$.fragment,c),at=i(c),K=u(c,"DIV",{class:!0});var wt=W(K);v(S.$$.fragment,wt),rt=i(wt),Ve=u(wt,"P",{"data-svelte-h":!0}),_(Ve)!=="svelte-1uhytbz"&&(Ve.textContent=It),ye=i(wt),Ge=u(wt,"P",{"data-svelte-h":!0}),_(Ge)!=="svelte-1pln4c3"&&(Ge.innerHTML=Yt),St=i(wt),At=u(wt,"P",{"data-svelte-h":!0}),_(At)!=="svelte-1be7e3c"&&(At.innerHTML=ft),He=i(wt),v(gt.$$.fragment,wt),Zt=i(wt),G=u(wt,"DIV",{class:!0});var Kt=W(G);v(ze.$$.fragment,Kt),ke=i(Kt),be=u(Kt,"P",{"data-svelte-h":!0}),_(be)!=="svelte-gdl8pp"&&(be.innerHTML=pn),Wt=i(Kt),v(we.$$.fragment,Kt),sn=i(Kt),v(D.$$.fragment,Kt),nn=i(Kt),v(pe.$$.fragment,Kt),Kt.forEach(a),wt.forEach(a),on=i(c),v(ee.$$.fragment,c),Ie=i(c),L=u(c,"DIV",{class:!0});var Ct=W(L);v(Ze.$$.fragment,Ct),Nt=i(Ct),_t=u(Ct,"P",{"data-svelte-h":!0}),_(_t)!=="svelte-1xscdal"&&(_t.textContent=te),qt=i(Ct),Rt=u(Ct,"P",{"data-svelte-h":!0}),_(Rt)!=="svelte-1pln4c3"&&(Rt.innerHTML=ge),Vt=i(Ct),Gt=u(Ct,"P",{"data-svelte-h":!0}),_(Gt)!=="svelte-1be7e3c"&&(Gt.innerHTML=it),mn=i(Ct),v(lt.$$.fragment,Ct),me=i(Ct),oe=u(Ct,"DIV",{class:!0});var dn=W(oe);v(Ce.$$.fragment,dn),bt=i(dn),De=u(dn,"P",{"data-svelte-h":!0}),_(De)!=="svelte-rma3bh"&&(De.innerHTML=an),se=i(dn),v(We.$$.fragment,dn),rn=i(dn),v(Q.$$.fragment,dn),dn.forEach(a),Ct.forEach(a),Tt=i(c),v(vt.$$.fragment,c),Le=i(c),ue=u(c,"DIV",{class:!0});var $t=W(ue);v(Mt.$$.fragment,$t),ne=i($t),Xe=u($t,"P",{"data-svelte-h":!0}),_(Xe)!=="svelte-6su72q"&&(Xe.textContent=Pt),Dt=i($t),Ht=u($t,"P",{"data-svelte-h":!0}),_(Ht)!=="svelte-1pln4c3"&&(Ht.innerHTML=yt),Oe=i($t),Ee=u($t,"P",{"data-svelte-h":!0}),_(Ee)!=="svelte-1be7e3c"&&(Ee.innerHTML=dt),Lt=i($t),v(h.$$.fragment,$t),s=i($t),$=u($t,"DIV",{class:!0});var en=W($);v(X.$$.fragment,en),N=i(en),Y=u(en,"P",{"data-svelte-h":!0}),_(Y)!=="svelte-1pg6lhr"&&(Y.innerHTML=ae),re=i(en),v(he.$$.fragment,en),O=i(en),v(A.$$.fragment,en),ie=i(en),v($e.$$.fragment,en),en.forEach(a),$t.forEach(a),_n=i(c),v(cn.$$.fragment,c),wn=i(c),Ke=u(c,"DIV",{class:!0});var Bt=W(Ke);v(bn.$$.fragment,Bt),$n=i(Bt),vn=u(Bt,"P",{"data-svelte-h":!0}),_(vn)!=="svelte-1iyf8cf"&&(vn.innerHTML=Zn),Bn=i(Bt),Mn=u(Bt,"P",{"data-svelte-h":!0}),_(Mn)!=="svelte-1pln4c3"&&(Mn.innerHTML=Wn),jn=i(Bt),yn=u(Bt,"P",{"data-svelte-h":!0}),_(yn)!=="svelte-1be7e3c"&&(yn.innerHTML=Nn),Jn=i(Bt),v(un.$$.fragment,Bt),Fn=i(Bt),Xt=u(Bt,"DIV",{class:!0});var tn=W(Xt);v(Tn.$$.fragment,tn),Un=i(tn),kn=u(tn,"P",{"data-svelte-h":!0}),_(kn)!=="svelte-1hfpkeh"&&(kn.innerHTML=qn),xn=i(tn),v(hn.$$.fragment,tn),zn=i(tn),v(fn.$$.fragment,tn),In=i(tn),v(gn.$$.fragment,tn),tn.forEach(a),Bt.forEach(a),this.h()},h(){Z(P,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(n,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(Me,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(H,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(G,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(K,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(oe,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(L,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z($,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ue,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(Xt,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(Ke,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8")},m(c,B){M(e,c,B),d(c,f,B),d(c,n,B),M(o,n,null),l(n,b),l(n,t),l(n,E),l(n,U),l(n,R),l(n,x),l(n,V),M(p,n,null),l(n,j),l(n,P),M(Te,P,null),l(P,jt),l(P,et),l(P,Jt),M(Ye,P,null),l(P,qe),M(tt,P,null),d(c,nt,B),M(le,c,B),d(c,Se,B),d(c,H,B),M(Je,H,null),l(H,Ft),l(H,Et),l(H,ct),l(H,ve),l(H,Fe),l(H,pt),l(H,Ut),M(st,H,null),l(H,Ae),l(H,Me),M(mt,Me,null),l(Me,Ue),l(Me,xt),l(Me,ce),M(I,Me,null),l(Me,ut),M(xe,Me,null),l(Me,Qt),M(ht,Me,null),d(c,Pe,B),M(Re,c,B),d(c,at,B),d(c,K,B),M(S,K,null),l(K,rt),l(K,Ve),l(K,ye),l(K,Ge),l(K,St),l(K,At),l(K,He),M(gt,K,null),l(K,Zt),l(K,G),M(ze,G,null),l(G,ke),l(G,be),l(G,Wt),M(we,G,null),l(G,sn),M(D,G,null),l(G,nn),M(pe,G,null),d(c,on,B),M(ee,c,B),d(c,Ie,B),d(c,L,B),M(Ze,L,null),l(L,Nt),l(L,_t),l(L,qt),l(L,Rt),l(L,Vt),l(L,Gt),l(L,mn),M(lt,L,null),l(L,me),l(L,oe),M(Ce,oe,null),l(oe,bt),l(oe,De),l(oe,se),M(We,oe,null),l(oe,rn),M(Q,oe,null),d(c,Tt,B),M(vt,c,B),d(c,Le,B),d(c,ue,B),M(Mt,ue,null),l(ue,ne),l(ue,Xe),l(ue,Dt),l(ue,Ht),l(ue,Oe),l(ue,Ee),l(ue,Lt),M(h,ue,null),l(ue,s),l(ue,$),M(X,$,null),l($,N),l($,Y),l($,re),M(he,$,null),l($,O),M(A,$,null),l($,ie),M($e,$,null),d(c,_n,B),M(cn,c,B),d(c,wn,B),d(c,Ke,B),M(bn,Ke,null),l(Ke,$n),l(Ke,vn),l(Ke,Bn),l(Ke,Mn),l(Ke,jn),l(Ke,yn),l(Ke,Jn),M(un,Ke,null),l(Ke,Fn),l(Ke,Xt),M(Tn,Xt,null),l(Xt,Un),l(Xt,kn),l(Xt,xn),M(hn,Xt,null),l(Xt,zn),M(fn,Xt,null),l(Xt,In),M(gn,Xt,null),Cn=!0},p(c,B){const ln={};B&2&&(ln.$$scope={dirty:B,ctx:c}),p.$set(ln);const kt={};B&2&&(kt.$$scope={dirty:B,ctx:c}),Ye.$set(kt);const Ot={};B&2&&(Ot.$$scope={dirty:B,ctx:c}),tt.$set(Ot);const wt={};B&2&&(wt.$$scope={dirty:B,ctx:c}),st.$set(wt);const Kt={};B&2&&(Kt.$$scope={dirty:B,ctx:c}),I.$set(Kt);const Ct={};B&2&&(Ct.$$scope={dirty:B,ctx:c}),xe.$set(Ct);const dn={};B&2&&(dn.$$scope={dirty:B,ctx:c}),ht.$set(dn);const $t={};B&2&&($t.$$scope={dirty:B,ctx:c}),gt.$set($t);const en={};B&2&&(en.$$scope={dirty:B,ctx:c}),we.$set(en);const Bt={};B&2&&(Bt.$$scope={dirty:B,ctx:c}),D.$set(Bt);const tn={};B&2&&(tn.$$scope={dirty:B,ctx:c}),pe.$set(tn);const Rn={};B&2&&(Rn.$$scope={dirty:B,ctx:c}),lt.$set(Rn);const Vn={};B&2&&(Vn.$$scope={dirty:B,ctx:c}),We.$set(Vn);const Gn={};B&2&&(Gn.$$scope={dirty:B,ctx:c}),Q.$set(Gn);const Hn={};B&2&&(Hn.$$scope={dirty:B,ctx:c}),h.$set(Hn);const Ln={};B&2&&(Ln.$$scope={dirty:B,ctx:c}),he.$set(Ln);const Xn={};B&2&&(Xn.$$scope={dirty:B,ctx:c}),A.$set(Xn);const En={};B&2&&(En.$$scope={dirty:B,ctx:c}),$e.$set(En);const Qn={};B&2&&(Qn.$$scope={dirty:B,ctx:c}),un.$set(Qn);const Yn={};B&2&&(Yn.$$scope={dirty:B,ctx:c}),hn.$set(Yn);const Sn={};B&2&&(Sn.$$scope={dirty:B,ctx:c}),fn.$set(Sn);const An={};B&2&&(An.$$scope={dirty:B,ctx:c}),gn.$set(An)},i(c){Cn\|\|(y(e.$$.fragment,c),y(o.$$.fragment,c),y(p.$$.fragment,c),y(Te.$$.fragment,c),y(Ye.$$.fragment,c),y(tt.$$.fragment,c),y(le.$$.fragment,c),y(Je.$$.fragment,c),y(st.$$.fragment,c),y(mt.$$.fragment,c),y(I.$$.fragment,c),y(xe.$$.fragment,c),y(ht.$$.fragment,c),y(Re.$$.fragment,c),y(S.$$.fragment,c),y(gt.$$.fragment,c),y(ze.$$.fragment,c),y(we.$$.fragment,c),y(D.$$.fragment,c),y(pe.$$.fragment,c),y(ee.$$.fragment,c),y(Ze.$$.fragment,c),y(lt.$$.fragment,c),y(Ce.$$.fragment,c),y(We.$$.fragment,c),y(Q.$$.fragment,c),y(vt.$$.fragment,c),y(Mt.$$.fragment,c),y(h.$$.fragment,c),y(X.$$.fragment,c),y(he.$$.fragment,c),y(A.$$.fragment,c),y($e.$$.fragment,c),y(cn.$$.fragment,c),y(bn.$$.fragment,c),y(un.$$.fragment,c),y(Tn.$$.fragment,c),y(hn.$$.fragment,c),y(fn.$$.fragment,c),y(gn.$$.fragment,c),Cn=!0)},o(c){k(e.$$.fragment,c),k(o.$$.fragment,c),k(p.$$.fragment,c),k(Te.$$.fragment,c),k(Ye.$$.fragment,c),k(tt.$$.fragment,c),k(le.$$.fragment,c),k(Je.$$.fragment,c),k(st.$$.fragment,c),k(mt.$$.fragment,c),k(I.$$.fragment,c),k(xe.$$.fragment,c),k(ht.$$.fragment,c),k(Re.$$.fragment,c),k(S.$$.fragment,c),k(gt.$$.fragment,c),k(ze.$$.fragment,c),k(we.$$.fragment,c),k(D.$$.fragment,c),k(pe.$$.fragment,c),k(ee.$$.fragment,c),k(Ze.$$.fragment,c),k(lt.$$.fragment,c),k(Ce.$$.fragment,c),k(We.$$.fragment,c),k(Q.$$.fragment,c),k(vt.$$.fragment,c),k(Mt.$$.fragment,c),k(h.$$.fragment,c),k(X.$$.fragment,c),k(he.$$.fragment,c),k(A.$$.fragment,c),k($e.$$.fragment,c),k(cn.$$.fragment,c),k(bn.$$.fragment,c),k(un.$$.fragment,c),k(Tn.$$.fragment,c),k(hn.$$.fragment,c),k(fn.$$.fragment,c),k(gn.$$.fragment,c),Cn=!1},d(c){c&&(a(f),a(n),a(nt),a(Se),a(H),a(Pe),a(at),a(K),a(on),a(Ie),a(L),a(Tt),a(Le),a(ue),a(_n),a(wn),a(Ke)),w(e,c),w(o),w(p),w(Te),w(Ye),w(tt),w(le,c),w(Je),w(st),w(mt),w(I),w(xe),w(ht),w(Re,c),w(S),w(gt),w(ze),w(we),w(D),w(pe),w(ee,c),w(Ze),w(lt),w(Ce),w(We),w(Q),w(vt,c),w(Mt),w(h),w(X),w(he),w(A),w($e),w(cn,c),w(bn),w(un),w(Tn),w(hn),w(fn),w(gn)}}}function Qo(C){let e,f;return e=new Pn({props:{$$slots:{default:[Eo]},$$scope:{ctx:C}}}),{c(){T(e.$$.fragment)},l(n){v(e.$$.fragment,n)},m(n,o){M(e,n,o),f=!0},p(n,o){const b={};o&2&&(b.$$scope={dirty:o,ctx:n}),e.$set(b)},i(n){f\|\|(y(e.$$.fragment,n),f=!0)},o(n){k(e.$$.fragment,n),f=!1},d(n){w(e,n)}}}function Yo(C){let e,f,n,o,b,t,g,E='<img alt="PyTorch" src="https://img.shields.io/badge/PyTorch-DE3412?style=flat&logo=pytorch&logoColor=white"/> <img alt="TensorFlow" src="https://img.shields.io/badge/TensorFlow-FF6F00?style=flat&logo=tensorflow&logoColor=white"/>',U,J,R,x,F=`The ConvBERT model was proposed in <a href="https://arxiv.org/abs/2008.02496" rel="nofollow">ConvBERT: Improving BERT with Span-based Dynamic Convolution</a> by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng
	Yan.`,V,p,j="The abstract from the paper is the following:",P,Te,jt=`<em>Pre-trained language models like BERT and its variants have recently achieved impressive performance in various
	natural language understanding tasks. However, BERT heavily relies on the global self-attention block and thus suffers
	large memory footprint and computation cost. Although all its attention heads query on the whole input sequence for
	generating the attention map from a global perspective, we observe some heads only need to learn local dependencies,
	which means the existence of computation redundancy. We therefore propose a novel span-based dynamic convolution to
	replace these self-attention heads to directly model local dependencies. The novel convolution heads, together with the
	rest self-attention heads, form a new mixed attention block that is more efficient at both global and local context
	learning. We equip BERT with this mixed attention design and build a ConvBERT model. Experiments have shown that
	ConvBERT significantly outperforms BERT and its variants in various downstream tasks, with lower training cost and
	fewer model parameters. Remarkably, ConvBERTbase model achieves 86.4 GLUE score, 0.7 higher than ELECTRAbase, while
	using less than 1/4 training cost. Code and pre-trained models will be released.</em>`,et,fe,Jt=`This model was contributed by <a href="https://huggingface.co/abhishek" rel="nofollow">abhishek</a>. The original implementation can be found
	here: <a href="https://github.com/yitu-opensource/ConvBert" rel="nofollow">https://github.com/yitu-opensource/ConvBert</a>`,Ye,qe,tt,nt,le='ConvBERT training tips are similar to those of BERT. For usage tips refer to <a href="bert">BERT documentation</a>.',Se,H,Je,Ft,Et='<li><a href="../tasks/sequence_classification">Text classification task guide</a></li> <li><a href="../tasks/token_classification">Token classification task guide</a></li> <li><a href="../tasks/question_answering">Question answering task guide</a></li> <li><a href="../tasks/masked_language_modeling">Masked language modeling task guide</a></li> <li><a href="../tasks/multiple_choice">Multiple choice task guide</a></li>',ot,ct,ve,_e,Fe,pt,de,Ut=`This is the configuration class to store the configuration of a <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertModel">ConvBertModel</a>. It is used to instantiate an
	ConvBERT model according to the specified arguments, defining the model architecture. Instantiating a configuration
	with the defaults will yield a similar configuration to that of the ConvBERT
	<a href="https://huggingface.co/YituTech/conv-bert-base" rel="nofollow">YituTech/conv-bert-base</a> architecture.`,st,Ae,Me=`Configuration objects inherit from <a href="/docs/transformers/pr_36839/en/main_classes/configuration#transformers.PretrainedConfig">PretrainedConfig</a> and can be used to control the model outputs. Read the
	documentation from <a href="/docs/transformers/pr_36839/en/main_classes/configuration#transformers.PretrainedConfig">PretrainedConfig</a> for more information.`,mt,Ue,xt,zt,ce,I,ut,xe,Qt,ht="Construct a ConvBERT tokenizer. Based on WordPiece.",Pe,Re,at=`This tokenizer inherits from <a href="/docs/transformers/pr_36839/en/main_classes/tokenizer#transformers.PreTrainedTokenizer">PreTrainedTokenizer</a> which contains most of the main methods. Users should refer to
	this superclass for more information regarding those methods.`,K,S,rt,Ve,It,ye=`Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
	adding special tokens. A ConvBERT sequence has the following format:`,Ge,Yt,St="<li>single sequence: <code>[CLS] X [SEP]</code></li> <li>pair of sequences: <code>[CLS] A [SEP] B [SEP]</code></li>",At,ft,He,gt,Zt,G=`Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
	special tokens using the tokenizer <code>prepare_for_model</code> method.`,ze,ke,be,pn,Wt,we="Create a mask from the two sequences passed to be used in a sequence-pair classification task. A ConvBERT sequence",sn,D,nn,pe,on="If <code>token_ids_1</code> is <code>None</code>, this method only returns the first portion of the mask (0s).",ee,Ie,L,Ze,Nt,_t,te,qt,Rt,ge,Vt="Construct a “fast” ConvBERT tokenizer (backed by HuggingFace’s <em>tokenizers</em> library). Based on WordPiece.",Gt,it,mn=`This tokenizer inherits from <a href="/docs/transformers/pr_36839/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast">PreTrainedTokenizerFast</a> which contains most of the main methods. Users should
	refer to this superclass for more information regarding those methods.`,lt,me,oe,Ce,bt,De=`Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and
	adding special tokens. A ConvBERT sequence has the following format:`,an,se,We="<li>single sequence: <code>[CLS] X [SEP]</code></li> <li>pair of sequences: <code>[CLS] A [SEP] B [SEP]</code></li>",rn,Q,Tt,vt,Le,ue="Create a mask from the two sequences passed to be used in a sequence-pair classification task. A ConvBERT sequence",Mt,ne,Xe,Pt,Dt="If <code>token_ids_1</code> is <code>None</code>, this method only returns the first portion of the mask (0s).",Ht,yt,Oe,Ee,dt,Lt,h;return b=new Ne({props:{title:"ConvBERT",local:"convbert",headingTag:"h1"}}),J=new Ne({props:{title:"Overview",local:"overview",headingTag:"h2"}}),qe=new Ne({props:{title:"Usage tips",local:"usage-tips",headingTag:"h2"}}),H=new Ne({props:{title:"Resources",local:"resources",headingTag:"h2"}}),ct=new Ne({props:{title:"ConvBertConfig",local:"transformers.ConvBertConfig",headingTag:"h2"}}),Fe=new q({props:{name:"class transformers.ConvBertConfig",anchor:"transformers.ConvBertConfig",parameters:[{name:"vocab_size",val:" = 30522"},{name:"hidden_size",val:" = 768"},{name:"num_hidden_layers",val:" = 12"},{name:"num_attention_heads",val:" = 12"},{name:"intermediate_size",val:" = 3072"},{name:"hidden_act",val:" = 'gelu'"},{name:"hidden_dropout_prob",val:" = 0.1"},{name:"attention_probs_dropout_prob",val:" = 0.1"},{name:"max_position_embeddings",val:" = 512"},{name:"type_vocab_size",val:" = 2"},{name:"initializer_range",val:" = 0.02"},{name:"layer_norm_eps",val:" = 1e-12"},{name:"pad_token_id",val:" = 1"},{name:"bos_token_id",val:" = 0"},{name:"eos_token_id",val:" = 2"},{name:"embedding_size",val:" = 768"},{name:"head_ratio",val:" = 2"},{name:"conv_kernel_size",val:" = 9"},{name:"num_groups",val:" = 1"},{name:"classifier_dropout",val:" = None"},{name:"**kwargs",val:""}],parametersDescription:[{anchor:"transformers.ConvBertConfig.vocab_size",description:`<strong>vocab_size</strong> (<code>int</code>, <em>optional</em>, defaults to 30522) —
	Vocabulary size of the ConvBERT model. Defines the number of different tokens that can be represented by
	the <code>inputs_ids</code> passed when calling <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertModel">ConvBertModel</a> or <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertModel">TFConvBertModel</a>.`,name:"vocab_size"},{anchor:"transformers.ConvBertConfig.hidden_size",description:`<strong>hidden_size</strong> (<code>int</code>, <em>optional</em>, defaults to 768) —
	Dimensionality of the encoder layers and the pooler layer.`,name:"hidden_size"},{anchor:"transformers.ConvBertConfig.num_hidden_layers",description:`<strong>num_hidden_layers</strong> (<code>int</code>, <em>optional</em>, defaults to 12) —
	Number of hidden layers in the Transformer encoder.`,name:"num_hidden_layers"},{anchor:"transformers.ConvBertConfig.num_attention_heads",description:`<strong>num_attention_heads</strong> (<code>int</code>, <em>optional</em>, defaults to 12) —
	Number of attention heads for each attention layer in the Transformer encoder.`,name:"num_attention_heads"},{anchor:"transformers.ConvBertConfig.intermediate_size",description:`<strong>intermediate_size</strong> (<code>int</code>, <em>optional</em>, defaults to 3072) —
	Dimensionality of the “intermediate” (i.e., feed-forward) layer in the Transformer encoder.`,name:"intermediate_size"},{anchor:"transformers.ConvBertConfig.hidden_act",description:`<strong>hidden_act</strong> (<code>str</code> or <code>function</code>, <em>optional</em>, defaults to <code>"gelu"</code>) —
	The non-linear activation function (function or string) in the encoder and pooler. If string, <code>"gelu"</code>,
	<code>"relu"</code>, <code>"selu"</code> and <code>"gelu_new"</code> are supported.`,name:"hidden_act"},{anchor:"transformers.ConvBertConfig.hidden_dropout_prob",description:`<strong>hidden_dropout_prob</strong> (<code>float</code>, <em>optional</em>, defaults to 0.1) —
	The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.`,name:"hidden_dropout_prob"},{anchor:"transformers.ConvBertConfig.attention_probs_dropout_prob",description:`<strong>attention_probs_dropout_prob</strong> (<code>float</code>, <em>optional</em>, defaults to 0.1) —
	The dropout ratio for the attention probabilities.`,name:"attention_probs_dropout_prob"},{anchor:"transformers.ConvBertConfig.max_position_embeddings",description:`<strong>max_position_embeddings</strong> (<code>int</code>, <em>optional</em>, defaults to 512) —
	The maximum sequence length that this model might ever be used with. Typically set this to something large
	just in case (e.g., 512 or 1024 or 2048).`,name:"max_position_embeddings"},{anchor:"transformers.ConvBertConfig.type_vocab_size",description:`<strong>type_vocab_size</strong> (<code>int</code>, <em>optional</em>, defaults to 2) —
	The vocabulary size of the <code>token_type_ids</code> passed when calling <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.ConvBertModel">ConvBertModel</a> or <a href="/docs/transformers/pr_36839/en/model_doc/convbert#transformers.TFConvBertModel">TFConvBertModel</a>.`,name:"type_vocab_size"},{anchor:"transformers.ConvBertConfig.initializer_range",description:`<strong>initializer_range</strong> (<code>float</code>, <em>optional</em>, defaults to 0.02) —
	The standard deviation of the truncated_normal_initializer for initializing all weight matrices.`,name:"initializer_range"},{anchor:"transformers.ConvBertConfig.layer_norm_eps",description:`<strong>layer_norm_eps</strong> (<code>float</code>, <em>optional</em>, defaults to 1e-12) —
	The epsilon used by the layer normalization layers.`,name:"layer_norm_eps"},{anchor:"transformers.ConvBertConfig.head_ratio",description:`<strong>head_ratio</strong> (<code>int</code>, <em>optional</em>, defaults to 2) —
	Ratio gamma to reduce the number of attention heads.`,name:"head_ratio"},{anchor:"transformers.ConvBertConfig.num_groups",description:`<strong>num_groups</strong> (<code>int</code>, <em>optional</em>, defaults to 1) —
	The number of groups for grouped linear layers for ConvBert model`,name:"num_groups"},{anchor:"transformers.ConvBertConfig.conv_kernel_size",description:`<strong>conv_kernel_size</strong> (<code>int</code>, <em>optional</em>, defaults to 9) —
	The size of the convolutional kernel.`,name:"conv_kernel_size"},{anchor:"transformers.ConvBertConfig.classifier_dropout",description:`<strong>classifier_dropout</strong> (<code>float</code>, <em>optional</em>) —
	The dropout ratio for the classification head.`,name:"classifier_dropout"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/configuration_convbert.py#L28"}}),Ue=new Be({props:{anchor:"transformers.ConvBertConfig.example",$$slots:{default:[so]},$$scope:{ctx:C}}}),zt=new Ne({props:{title:"ConvBertTokenizer",local:"transformers.ConvBertTokenizer",headingTag:"h2"}}),ut=new q({props:{name:"class transformers.ConvBertTokenizer",anchor:"transformers.ConvBertTokenizer",parameters:[{name:"vocab_file",val:""},{name:"do_lower_case",val:" = True"},{name:"do_basic_tokenize",val:" = True"},{name:"never_split",val:" = None"},{name:"unk_token",val:" = '[UNK]'"},{name:"sep_token",val:" = '[SEP]'"},{name:"pad_token",val:" = '[PAD]'"},{name:"cls_token",val:" = '[CLS]'"},{name:"mask_token",val:" = '[MASK]'"},{name:"tokenize_chinese_chars",val:" = True"},{name:"strip_accents",val:" = None"},{name:"clean_up_tokenization_spaces",val:" = True"},{name:"**kwargs",val:""}],parametersDescription:[{anchor:"transformers.ConvBertTokenizer.vocab_file",description:`<strong>vocab_file</strong> (<code>str</code>) —
	File containing the vocabulary.`,name:"vocab_file"},{anchor:"transformers.ConvBertTokenizer.do_lower_case",description:`<strong>do_lower_case</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) —
	Whether or not to lowercase the input when tokenizing.`,name:"do_lower_case"},{anchor:"transformers.ConvBertTokenizer.do_basic_tokenize",description:`<strong>do_basic_tokenize</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) —
	Whether or not to do basic tokenization before WordPiece.`,name:"do_basic_tokenize"},{anchor:"transformers.ConvBertTokenizer.never_split",description:`<strong>never_split</strong> (<code>Iterable</code>, <em>optional</em>) —
	Collection of tokens which will never be split during tokenization. Only has an effect when
	<code>do_basic_tokenize=True</code>`,name:"never_split"},{anchor:"transformers.ConvBertTokenizer.unk_token",description:`<strong>unk_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[UNK]"</code>) —
	The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
	token instead.`,name:"unk_token"},{anchor:"transformers.ConvBertTokenizer.sep_token",description:`<strong>sep_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[SEP]"</code>) —
	The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
	sequence classification or for a text and a question for question answering. It is also used as the last
	token of a sequence built with special tokens.`,name:"sep_token"},{anchor:"transformers.ConvBertTokenizer.pad_token",description:`<strong>pad_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[PAD]"</code>) —
	The token used for padding, for example when batching sequences of different lengths.`,name:"pad_token"},{anchor:"transformers.ConvBertTokenizer.cls_token",description:`<strong>cls_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[CLS]"</code>) —
	The classifier token which is used when doing sequence classification (classification of the whole sequence
	instead of per-token classification). It is the first token of the sequence when built with special tokens.`,name:"cls_token"},{anchor:"transformers.ConvBertTokenizer.mask_token",description:`<strong>mask_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[MASK]"</code>) —
	The token used for masking values. This is the token used when training this model with masked language
	modeling. This is the token which the model will try to predict.`,name:"mask_token"},{anchor:"transformers.ConvBertTokenizer.tokenize_chinese_chars",description:`<strong>tokenize_chinese_chars</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) —
	Whether or not to tokenize Chinese characters.</p>
	<p>This should likely be deactivated for Japanese (see this
	<a href="https://github.com/huggingface/transformers/issues/328" rel="nofollow">issue</a>).`,name:"tokenize_chinese_chars"},{anchor:"transformers.ConvBertTokenizer.strip_accents",description:`<strong>strip_accents</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to strip all accents. If this option is not specified, then it will be determined by the
	value for <code>lowercase</code> (as in the original ConvBERT).`,name:"strip_accents"},{anchor:"transformers.ConvBertTokenizer.clean_up_tokenization_spaces",description:`<strong>clean_up_tokenization_spaces</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) —
	Whether or not to cleanup spaces after decoding, cleanup consists in removing potential artifacts like
	extra spaces.`,name:"clean_up_tokenization_spaces"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert.py#L54"}}),rt=new q({props:{name:"build_inputs_with_special_tokens",anchor:"transformers.ConvBertTokenizer.build_inputs_with_special_tokens",parameters:[{name:"token_ids_0",val:": typing.List[int]"},{name:"token_ids_1",val:": typing.Optional[typing.List[int]] = None"}],parametersDescription:[{anchor:"transformers.ConvBertTokenizer.build_inputs_with_special_tokens.token_ids_0",description:`<strong>token_ids_0</strong> (<code>List[int]</code>) —
	List of IDs to which the special tokens will be added.`,name:"token_ids_0"},{anchor:"transformers.ConvBertTokenizer.build_inputs_with_special_tokens.token_ids_1",description:`<strong>token_ids_1</strong> (<code>List[int]</code>, <em>optional</em>) —
	Optional second list of IDs for sequence pairs.`,name:"token_ids_1"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert.py#L189",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>List of <a href="../glossary#input-ids">input IDs</a> with the appropriate special tokens.</p>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><code>List[int]</code></p>
	`}}),He=new q({props:{name:"get_special_tokens_mask",anchor:"transformers.ConvBertTokenizer.get_special_tokens_mask",parameters:[{name:"token_ids_0",val:": typing.List[int]"},{name:"token_ids_1",val:": typing.Optional[typing.List[int]] = None"},{name:"already_has_special_tokens",val:": bool = False"}],parametersDescription:[{anchor:"transformers.ConvBertTokenizer.get_special_tokens_mask.token_ids_0",description:`<strong>token_ids_0</strong> (<code>List[int]</code>) —
	List of IDs.`,name:"token_ids_0"},{anchor:"transformers.ConvBertTokenizer.get_special_tokens_mask.token_ids_1",description:`<strong>token_ids_1</strong> (<code>List[int]</code>, <em>optional</em>) —
	Optional second list of IDs for sequence pairs.`,name:"token_ids_1"},{anchor:"transformers.ConvBertTokenizer.get_special_tokens_mask.already_has_special_tokens",description:`<strong>already_has_special_tokens</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>False</code>) —
	Whether or not the token list is already formatted with special tokens for the model.`,name:"already_has_special_tokens"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert.py#L214",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.</p>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><code>List[int]</code></p>
	`}}),be=new q({props:{name:"create_token_type_ids_from_sequences",anchor:"transformers.ConvBertTokenizer.create_token_type_ids_from_sequences",parameters:[{name:"token_ids_0",val:": typing.List[int]"},{name:"token_ids_1",val:": typing.Optional[typing.List[int]] = None"}],parametersDescription:[{anchor:"transformers.ConvBertTokenizer.create_token_type_ids_from_sequences.token_ids_0",description:`<strong>token_ids_0</strong> (<code>List[int]</code>) —
	List of IDs.`,name:"token_ids_0"},{anchor:"transformers.ConvBertTokenizer.create_token_type_ids_from_sequences.token_ids_1",description:`<strong>token_ids_1</strong> (<code>List[int]</code>, <em>optional</em>) —
	Optional second list of IDs for sequence pairs.`,name:"token_ids_1"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert.py#L242",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>List of <a href="../glossary#token-type-ids">token type IDs</a> according to the given sequence(s).</p>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><code>List[int]</code></p>
	`}}),D=new Be({props:{anchor:"transformers.ConvBertTokenizer.create_token_type_ids_from_sequences.example",$$slots:{default:[ao]},$$scope:{ctx:C}}}),L=new q({props:{name:"save_vocabulary",anchor:"transformers.ConvBertTokenizer.save_vocabulary",parameters:[{name:"save_directory",val:": str"},{name:"filename_prefix",val:": typing.Optional[str] = None"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert.py#L271"}}),Nt=new Ne({props:{title:"ConvBertTokenizerFast",local:"transformers.ConvBertTokenizerFast",headingTag:"h2"}}),qt=new q({props:{name:"class transformers.ConvBertTokenizerFast",anchor:"transformers.ConvBertTokenizerFast",parameters:[{name:"vocab_file",val:" = None"},{name:"tokenizer_file",val:" = None"},{name:"do_lower_case",val:" = True"},{name:"unk_token",val:" = '[UNK]'"},{name:"sep_token",val:" = '[SEP]'"},{name:"pad_token",val:" = '[PAD]'"},{name:"cls_token",val:" = '[CLS]'"},{name:"mask_token",val:" = '[MASK]'"},{name:"tokenize_chinese_chars",val:" = True"},{name:"strip_accents",val:" = None"},{name:"**kwargs",val:""}],parametersDescription:[{anchor:"transformers.ConvBertTokenizerFast.vocab_file",description:`<strong>vocab_file</strong> (<code>str</code>) —
	File containing the vocabulary.`,name:"vocab_file"},{anchor:"transformers.ConvBertTokenizerFast.do_lower_case",description:`<strong>do_lower_case</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) —
	Whether or not to lowercase the input when tokenizing.`,name:"do_lower_case"},{anchor:"transformers.ConvBertTokenizerFast.unk_token",description:`<strong>unk_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[UNK]"</code>) —
	The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
	token instead.`,name:"unk_token"},{anchor:"transformers.ConvBertTokenizerFast.sep_token",description:`<strong>sep_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[SEP]"</code>) —
	The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
	sequence classification or for a text and a question for question answering. It is also used as the last
	token of a sequence built with special tokens.`,name:"sep_token"},{anchor:"transformers.ConvBertTokenizerFast.pad_token",description:`<strong>pad_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[PAD]"</code>) —
	The token used for padding, for example when batching sequences of different lengths.`,name:"pad_token"},{anchor:"transformers.ConvBertTokenizerFast.cls_token",description:`<strong>cls_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[CLS]"</code>) —
	The classifier token which is used when doing sequence classification (classification of the whole sequence
	instead of per-token classification). It is the first token of the sequence when built with special tokens.`,name:"cls_token"},{anchor:"transformers.ConvBertTokenizerFast.mask_token",description:`<strong>mask_token</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"[MASK]"</code>) —
	The token used for masking values. This is the token used when training this model with masked language
	modeling. This is the token which the model will try to predict.`,name:"mask_token"},{anchor:"transformers.ConvBertTokenizerFast.clean_text",description:`<strong>clean_text</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) —
	Whether or not to clean the text before tokenization by removing any control characters and replacing all
	whitespaces by the classic one.`,name:"clean_text"},{anchor:"transformers.ConvBertTokenizerFast.tokenize_chinese_chars",description:`<strong>tokenize_chinese_chars</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) —
	Whether or not to tokenize Chinese characters. This should likely be deactivated for Japanese (see <a href="https://github.com/huggingface/transformers/issues/328" rel="nofollow">this
	issue</a>).`,name:"tokenize_chinese_chars"},{anchor:"transformers.ConvBertTokenizerFast.strip_accents",description:`<strong>strip_accents</strong> (<code>bool</code>, <em>optional</em>) —
	Whether or not to strip all accents. If this option is not specified, then it will be determined by the
	value for <code>lowercase</code> (as in the original ConvBERT).`,name:"strip_accents"},{anchor:"transformers.ConvBertTokenizerFast.wordpieces_prefix",description:`<strong>wordpieces_prefix</strong> (<code>str</code>, <em>optional</em>, defaults to <code>"##"</code>) —
	The prefix for subwords.`,name:"wordpieces_prefix"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert_fast.py#L33"}}),oe=new q({props:{name:"build_inputs_with_special_tokens",anchor:"transformers.ConvBertTokenizerFast.build_inputs_with_special_tokens",parameters:[{name:"token_ids_0",val:""},{name:"token_ids_1",val:" = None"}],parametersDescription:[{anchor:"transformers.ConvBertTokenizerFast.build_inputs_with_special_tokens.token_ids_0",description:`<strong>token_ids_0</strong> (<code>List[int]</code>) —
	List of IDs to which the special tokens will be added.`,name:"token_ids_0"},{anchor:"transformers.ConvBertTokenizerFast.build_inputs_with_special_tokens.token_ids_1",description:`<strong>token_ids_1</strong> (<code>List[int]</code>, <em>optional</em>) —
	Optional second list of IDs for sequence pairs.`,name:"token_ids_1"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert_fast.py#L118",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>List of <a href="../glossary#input-ids">input IDs</a> with the appropriate special tokens.</p>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><code>List[int]</code></p>
	`}}),Tt=new q({props:{name:"create_token_type_ids_from_sequences",anchor:"transformers.ConvBertTokenizerFast.create_token_type_ids_from_sequences",parameters:[{name:"token_ids_0",val:": typing.List[int]"},{name:"token_ids_1",val:": typing.Optional[typing.List[int]] = None"}],parametersDescription:[{anchor:"transformers.ConvBertTokenizerFast.create_token_type_ids_from_sequences.token_ids_0",description:`<strong>token_ids_0</strong> (<code>List[int]</code>) —
	List of IDs.`,name:"token_ids_0"},{anchor:"transformers.ConvBertTokenizerFast.create_token_type_ids_from_sequences.token_ids_1",description:`<strong>token_ids_1</strong> (<code>List[int]</code>, <em>optional</em>) —
	Optional second list of IDs for sequence pairs.`,name:"token_ids_1"}],source:"https://github.com/huggingface/transformers/blob/vr_36839/src/transformers/models/convbert/tokenization_convbert_fast.py#L142",returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>


	<p>List of <a href="../glossary#token-type-ids">token type IDs</a> according to the given sequence(s).</p>
	`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>


	<p><code>List[int]</code></p>
	`}}),ne=new Be({props:{anchor:"transformers.ConvBertTokenizerFast.create_token_type_ids_from_sequences.example",$$slots:{default:[ro]},$$scope:{ctx:C}}}),yt=new no({props:{pytorch:!0,tensorflow:!0,jax:!1,$$slots:{tensorflow:[Qo],pytorch:[yo]},$$scope:{ctx:C}}}),Ee=new oo({props:{source:"https://github.com/huggingface/transformers/blob/main/docs/source/en/model_doc/convbert.md"}}),{c(){e=m("meta"),f=r(),n=m("p"),o=r(),T(b.$$.fragment),t=r(),g=m("div"),g.innerHTML=E,U=r(),T(J.$$.fragment),R=r(),x=m("p"),x.innerHTML=F,V=r(),p=m("p"),p.textContent=j,P=r(),Te=m("p"),Te.innerHTML=jt,et=r(),fe=m("p"),fe.innerHTML=Jt,Ye=r(),T(qe.$$.fragment),tt=r(),nt=m("p"),nt.innerHTML=le,Se=r(),T(H.$$.fragment),Je=r(),Ft=m("ul"),Ft.innerHTML=Et,ot=r(),T(ct.$$.fragment),ve=r(),_e=m("div"),T(Fe.$$.fragment),pt=r(),de=m("p"),de.innerHTML=Ut,st=r(),Ae=m("p"),Ae.innerHTML=Me,mt=r(),T(Ue.$$.fragment),xt=r(),T(zt.$$.fragment),ce=r(),I=m("div"),T(ut.$$.fragment),xe=r(),Qt=m("p"),Qt.textContent=ht,Pe=r(),Re=m("p"),Re.innerHTML=at,K=r(),S=m("div"),T(rt.$$.fragment),Ve=r(),It=m("p"),It.textContent=ye,Ge=r(),Yt=m("ul"),Yt.innerHTML=St,At=r(),ft=m("div"),T(He.$$.fragment),gt=r(),Zt=m("p"),Zt.innerHTML=G,ze=r(),ke=m("div"),T(be.$$.fragment),pn=r(),Wt=m("p"),Wt.textContent=we,sn=r(),T(D.$$.fragment),nn=r(),pe=m("p"),pe.innerHTML=on,ee=r(),Ie=m("div"),T(L.$$.fragment),Ze=r(),T(Nt.$$.fragment),_t=r(),te=m("div"),T(qt.$$.fragment),Rt=r(),ge=m("p"),ge.innerHTML=Vt,Gt=r(),it=m("p"),it.innerHTML=mn,lt=r(),me=m("div"),T(oe.$$.fragment),Ce=r(),bt=m("p"),bt.textContent=De,an=r(),se=m("ul"),se.innerHTML=We,rn=r(),Q=m("div"),T(Tt.$$.fragment),vt=r(),Le=m("p"),Le.textContent=ue,Mt=r(),T(ne.$$.fragment),Xe=r(),Pt=m("p"),Pt.innerHTML=Dt,Ht=r(),T(yt.$$.fragment),Oe=r(),T(Ee.$$.fragment),dt=r(),Lt=m("p"),this.h()},l(s){const $=to("svelte-u9bgzb",document.head);e=u($,"META",{name:!0,content:!0}),$.forEach(a),f=i(s),n=u(s,"P",{}),W(n).forEach(a),o=i(s),v(b.$$.fragment,s),t=i(s),g=u(s,"DIV",{class:!0,"data-svelte-h":!0}),_(g)!=="svelte-ltwe0m"&&(g.innerHTML=E),U=i(s),v(J.$$.fragment,s),R=i(s),x=u(s,"P",{"data-svelte-h":!0}),_(x)!=="svelte-91mste"&&(x.innerHTML=F),V=i(s),p=u(s,"P",{"data-svelte-h":!0}),_(p)!=="svelte-vfdo9a"&&(p.textContent=j),P=i(s),Te=u(s,"P",{"data-svelte-h":!0}),_(Te)!=="svelte-75os3a"&&(Te.innerHTML=jt),et=i(s),fe=u(s,"P",{"data-svelte-h":!0}),_(fe)!=="svelte-18ok2q9"&&(fe.innerHTML=Jt),Ye=i(s),v(qe.$$.fragment,s),tt=i(s),nt=u(s,"P",{"data-svelte-h":!0}),_(nt)!=="svelte-mpj4bu"&&(nt.innerHTML=le),Se=i(s),v(H.$$.fragment,s),Je=i(s),Ft=u(s,"UL",{"data-svelte-h":!0}),_(Ft)!=="svelte-mgusi3"&&(Ft.innerHTML=Et),ot=i(s),v(ct.$$.fragment,s),ve=i(s),_e=u(s,"DIV",{class:!0});var X=W(_e);v(Fe.$$.fragment,X),pt=i(X),de=u(X,"P",{"data-svelte-h":!0}),_(de)!=="svelte-v5j4o9"&&(de.innerHTML=Ut),st=i(X),Ae=u(X,"P",{"data-svelte-h":!0}),_(Ae)!=="svelte-4potwj"&&(Ae.innerHTML=Me),mt=i(X),v(Ue.$$.fragment,X),X.forEach(a),xt=i(s),v(zt.$$.fragment,s),ce=i(s),I=u(s,"DIV",{class:!0});var N=W(I);v(ut.$$.fragment,N),xe=i(N),Qt=u(N,"P",{"data-svelte-h":!0}),_(Qt)!=="svelte-3unw68"&&(Qt.textContent=ht),Pe=i(N),Re=u(N,"P",{"data-svelte-h":!0}),_(Re)!=="svelte-hglat1"&&(Re.innerHTML=at),K=i(N),S=u(N,"DIV",{class:!0});var Y=W(S);v(rt.$$.fragment,Y),Ve=i(Y),It=u(Y,"P",{"data-svelte-h":!0}),_(It)!=="svelte-d1d9j4"&&(It.textContent=ye),Ge=i(Y),Yt=u(Y,"UL",{"data-svelte-h":!0}),_(Yt)!=="svelte-xi6653"&&(Yt.innerHTML=St),Y.forEach(a),At=i(N),ft=u(N,"DIV",{class:!0});var ae=W(ft);v(He.$$.fragment,ae),gt=i(ae),Zt=u(ae,"P",{"data-svelte-h":!0}),_(Zt)!=="svelte-1f4f5kp"&&(Zt.innerHTML=G),ae.forEach(a),ze=i(N),ke=u(N,"DIV",{class:!0});var re=W(ke);v(be.$$.fragment,re),pn=i(re),Wt=u(re,"P",{"data-svelte-h":!0}),_(Wt)!=="svelte-y3zhit"&&(Wt.textContent=we),sn=i(re),v(D.$$.fragment,re),nn=i(re),pe=u(re,"P",{"data-svelte-h":!0}),_(pe)!=="svelte-owoxgn"&&(pe.innerHTML=on),re.forEach(a),ee=i(N),Ie=u(N,"DIV",{class:!0});var he=W(Ie);v(L.$$.fragment,he),he.forEach(a),N.forEach(a),Ze=i(s),v(Nt.$$.fragment,s),_t=i(s),te=u(s,"DIV",{class:!0});var O=W(te);v(qt.$$.fragment,O),Rt=i(O),ge=u(O,"P",{"data-svelte-h":!0}),_(ge)!=="svelte-1u3sh1x"&&(ge.innerHTML=Vt),Gt=i(O),it=u(O,"P",{"data-svelte-h":!0}),_(it)!=="svelte-a9mnn5"&&(it.innerHTML=mn),lt=i(O),me=u(O,"DIV",{class:!0});var A=W(me);v(oe.$$.fragment,A),Ce=i(A),bt=u(A,"P",{"data-svelte-h":!0}),_(bt)!=="svelte-d1d9j4"&&(bt.textContent=De),an=i(A),se=u(A,"UL",{"data-svelte-h":!0}),_(se)!=="svelte-xi6653"&&(se.innerHTML=We),A.forEach(a),rn=i(O),Q=u(O,"DIV",{class:!0});var ie=W(Q);v(Tt.$$.fragment,ie),vt=i(ie),Le=u(ie,"P",{"data-svelte-h":!0}),_(Le)!=="svelte-y3zhit"&&(Le.textContent=ue),Mt=i(ie),v(ne.$$.fragment,ie),Xe=i(ie),Pt=u(ie,"P",{"data-svelte-h":!0}),_(Pt)!=="svelte-owoxgn"&&(Pt.innerHTML=Dt),ie.forEach(a),O.forEach(a),Ht=i(s),v(yt.$$.fragment,s),Oe=i(s),v(Ee.$$.fragment,s),dt=i(s),Lt=u(s,"P",{}),W(Lt).forEach(a),this.h()},h(){Z(e,"name","hf:doc:metadata"),Z(e,"content",So),Z(g,"class","flex flex-wrap space-x-1"),Z(_e,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(S,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ft,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(ke,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(Ie,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(I,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(me,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(Q,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"),Z(te,"class","docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8")},m(s,$){l(document.head,e),d(s,f,$),d(s,n,$),d(s,o,$),M(b,s,$),d(s,t,$),d(s,g,$),d(s,U,$),M(J,s,$),d(s,R,$),d(s,x,$),d(s,V,$),d(s,p,$),d(s,P,$),d(s,Te,$),d(s,et,$),d(s,fe,$),d(s,Ye,$),M(qe,s,$),d(s,tt,$),d(s,nt,$),d(s,Se,$),M(H,s,$),d(s,Je,$),d(s,Ft,$),d(s,ot,$),M(ct,s,$),d(s,ve,$),d(s,_e,$),M(Fe,_e,null),l(_e,pt),l(_e,de),l(_e,st),l(_e,Ae),l(_e,mt),M(Ue,_e,null),d(s,xt,$),M(zt,s,$),d(s,ce,$),d(s,I,$),M(ut,I,null),l(I,xe),l(I,Qt),l(I,Pe),l(I,Re),l(I,K),l(I,S),M(rt,S,null),l(S,Ve),l(S,It),l(S,Ge),l(S,Yt),l(I,At),l(I,ft),M(He,ft,null),l(ft,gt),l(ft,Zt),l(I,ze),l(I,ke),M(be,ke,null),l(ke,pn),l(ke,Wt),l(ke,sn),M(D,ke,null),l(ke,nn),l(ke,pe),l(I,ee),l(I,Ie),M(L,Ie,null),d(s,Ze,$),M(Nt,s,$),d(s,_t,$),d(s,te,$),M(qt,te,null),l(te,Rt),l(te,ge),l(te,Gt),l(te,it),l(te,lt),l(te,me),M(oe,me,null),l(me,Ce),l(me,bt),l(me,an),l(me,se),l(te,rn),l(te,Q),M(Tt,Q,null),l(Q,vt),l(Q,Le),l(Q,Mt),M(ne,Q,null),l(Q,Xe),l(Q,Pt),d(s,Ht,$),M(yt,s,$),d(s,Oe,$),M(Ee,s,$),d(s,dt,$),d(s,Lt,$),h=!0},p(s,[$]){const X={};$&2&&(X.$$scope={dirty:$,ctx:s}),Ue.$set(X);const N={};$&2&&(N.$$scope={dirty:$,ctx:s}),D.$set(N);const Y={};$&2&&(Y.$$scope={dirty:$,ctx:s}),ne.$set(Y);const ae={};$&2&&(ae.$$scope={dirty:$,ctx:s}),yt.$set(ae)},i(s){h\|\|(y(b.$$.fragment,s),y(J.$$.fragment,s),y(qe.$$.fragment,s),y(H.$$.fragment,s),y(ct.$$.fragment,s),y(Fe.$$.fragment,s),y(Ue.$$.fragment,s),y(zt.$$.fragment,s),y(ut.$$.fragment,s),y(rt.$$.fragment,s),y(He.$$.fragment,s),y(be.$$.fragment,s),y(D.$$.fragment,s),y(L.$$.fragment,s),y(Nt.$$.fragment,s),y(qt.$$.fragment,s),y(oe.$$.fragment,s),y(Tt.$$.fragment,s),y(ne.$$.fragment,s),y(yt.$$.fragment,s),y(Ee.$$.fragment,s),h=!0)},o(s){k(b.$$.fragment,s),k(J.$$.fragment,s),k(qe.$$.fragment,s),k(H.$$.fragment,s),k(ct.$$.fragment,s),k(Fe.$$.fragment,s),k(Ue.$$.fragment,s),k(zt.$$.fragment,s),k(ut.$$.fragment,s),k(rt.$$.fragment,s),k(He.$$.fragment,s),k(be.$$.fragment,s),k(D.$$.fragment,s),k(L.$$.fragment,s),k(Nt.$$.fragment,s),k(qt.$$.fragment,s),k(oe.$$.fragment,s),k(Tt.$$.fragment,s),k(ne.$$.fragment,s),k(yt.$$.fragment,s),k(Ee.$$.fragment,s),h=!1},d(s){s&&(a(f),a(n),a(o),a(t),a(g),a(U),a(R),a(x),a(V),a(p),a(P),a(Te),a(et),a(fe),a(Ye),a(tt),a(nt),a(Se),a(Je),a(Ft),a(ot),a(ve),a(_e),a(xt),a(ce),a(I),a(Ze),a(_t),a(te),a(Ht),a(Oe),a(dt),a(Lt)),a(e),w(b,s),w(J,s),w(qe,s),w(H,s),w(ct,s),w(Fe),w(Ue),w(zt,s),w(ut),w(rt),w(He),w(be),w(D),w(L),w(Nt,s),w(qt),w(oe),w(Tt),w(ne),w(yt,s),w(Ee,s)}}}const So='{"title":"ConvBERT","local":"convbert","sections":[{"title":"Overview","local":"overview","sections":[],"depth":2},{"title":"Usage tips","local":"usage-tips","sections":[],"depth":2},{"title":"Resources","local":"resources","sections":[],"depth":2},{"title":"ConvBertConfig","local":"transformers.ConvBertConfig","sections":[],"depth":2},{"title":"ConvBertTokenizer","local":"transformers.ConvBertTokenizer","sections":[],"depth":2},{"title":"ConvBertTokenizerFast","local":"transformers.ConvBertTokenizerFast","sections":[],"depth":2},{"title":"ConvBertModel","local":"transformers.ConvBertModel","sections":[],"depth":2},{"title":"ConvBertForMaskedLM","local":"transformers.ConvBertForMaskedLM","sections":[],"depth":2},{"title":"ConvBertForSequenceClassification","local":"transformers.ConvBertForSequenceClassification","sections":[],"depth":2},{"title":"ConvBertForMultipleChoice","local":"transformers.ConvBertForMultipleChoice","sections":[],"depth":2},{"title":"ConvBertForTokenClassification","local":"transformers.ConvBertForTokenClassification","sections":[],"depth":2},{"title":"ConvBertForQuestionAnswering","local":"transformers.ConvBertForQuestionAnswering","sections":[],"depth":2},{"title":"TFConvBertModel","local":"transformers.TFConvBertModel","sections":[],"depth":2},{"title":"TFConvBertForMaskedLM","local":"transformers.TFConvBertForMaskedLM","sections":[],"depth":2},{"title":"TFConvBertForSequenceClassification","local":"transformers.TFConvBertForSequenceClassification","sections":[],"depth":2},{"title":"TFConvBertForMultipleChoice","local":"transformers.TFConvBertForMultipleChoice","sections":[],"depth":2},{"title":"TFConvBertForTokenClassification","local":"transformers.TFConvBertForTokenClassification","sections":[],"depth":2},{"title":"TFConvBertForQuestionAnswering","local":"transformers.TFConvBertForQuestionAnswering","sections":[],"depth":2}],"depth":1}';function Ao(C){return On(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class ss extends Kn{constructor(e){super(),eo(this,e,Ao,Yo,Dn,{})}}export{ss as component};

Xet Storage Details

Size:: 239 kB
Xet hash:: 58d3002ff7bcdfb2aa03c14a890ac4ade63aecbd919945baabce4296857ff579

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