import{s as P,n as O,o as ss}from"../chunks/scheduler.7da89386.js";import{S as ls,i as es,g as T,s as t,r as j,A as as,h as b,f as e,c as n,j as K,u as U,x as v,k as D,y as ts,a,v as w,d as Z,t as W,w as k}from"../chunks/index.20910acc.js";import{C as x}from"../chunks/CodeBlock.143bd81e.js";import{H as S,E as ns}from"../chunks/getInferenceSnippets.7cf363b6.js";function rs(q){let r,V,G,f,p,I,c,Q="To run the 🤗 Transformers examples make sure you have installed the following libraries:",X,M,R,i,_,o,A='The metrics in <code>evaluate</code> can be easily integrated with the <a href="https://huggingface.co/docs/transformers/v4.56.2/en/main_classes/trainer#transformers.Trainer" rel="nofollow">Trainer</a>. The <code>Trainer</code> accepts a <code>compute_metrics</code> keyword argument that passes a function to compute metrics. One can specify the evaluation interval with <code>evaluation_strategy</code> in the <code>TrainerArguments</code>, and based on that, the model is evaluated accordingly, and the predictions and labels passed to <code>compute_metrics</code>.',C,d,E,m,F,y,$='We can use the <a href="https://huggingface.co/docs/transformers/v4.56.2/en/main_classes/trainer#transformers.Seq2SeqTrainer" rel="nofollow">Seq2SeqTrainer</a> for sequence-to-sequence tasks such as translation or summarization. For such generative tasks usually metrics such as ROUGE or BLEU are evaluated. However, these metrics require that we generate some text with the model rather than a single forward pass as with e.g. classification. The <code>Seq2SeqTrainer</code> allows for the use of the generate method when setting <code>predict_with_generate=True</code> which will generate text for each sample in the evaluation set. That means we evaluate generated text within the <code>compute_metric</code> function. We just need to decode the predictions and labels first.',g,J,z,u,L="You can use any <code>evaluate</code> metric with the <code>Trainer</code> and <code>Seq2SeqTrainer</code> as long as they are compatible with the task and predictions. In case you don’t want to train a model but just evaluate an existing model you can replace <code>trainer.train()</code> with <code>trainer.evaluate()</code> in the above scripts.",Y,h,N,B,H;return p=new S({props:{title:"🤗 Transformers",local:"-transformers",headingTag:"h1"}}),M=new x({props:{code:"cGlwJTIwaW5zdGFsbCUyMGRhdGFzZXRzJTIwdHJhbnNmb3JtZXJzJTIwdG9yY2glMjBldmFsdWF0ZSUyMG5sdGslMjByb3VnZV9zY29yZQ==",highlighted:"pip install datasets transformers torch evaluate nltk rouge_score",wrap:!1}}),i=new S({props:{title:"Trainer",local:"trainer",headingTag:"h2"}}),d=new x({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, AutoModelForSequenceClassification, TrainingArguments, Trainer
<span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
<span class="hljs-keyword">import</span> evaluate

<span class="hljs-comment"># Prepare and tokenize dataset</span>
dataset = load_dataset(<span class="hljs-string">&quot;yelp_review_full&quot;</span>)
tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">&quot;bert-base-cased&quot;</span>)

<span class="hljs-keyword">def</span> <span class="hljs-title function_">tokenize_function</span>(<span class="hljs-params">examples</span>):
    <span class="hljs-keyword">return</span> tokenizer(examples[<span class="hljs-string">&quot;text&quot;</span>], padding=<span class="hljs-string">&quot;max_length&quot;</span>, truncation=<span class="hljs-literal">True</span>)

tokenized_datasets = dataset.<span class="hljs-built_in">map</span>(tokenize_function, batched=<span class="hljs-literal">True</span>)

small_train_dataset = tokenized_datasets[<span class="hljs-string">&quot;train&quot;</span>].shuffle(seed=<span class="hljs-number">42</span>).select(<span class="hljs-built_in">range</span>(<span class="hljs-number">200</span>))
small_eval_dataset = tokenized_datasets[<span class="hljs-string">&quot;test&quot;</span>].shuffle(seed=<span class="hljs-number">42</span>).select(<span class="hljs-built_in">range</span>(<span class="hljs-number">200</span>))

<span class="hljs-comment"># Setup evaluation </span>
metric = evaluate.load(<span class="hljs-string">&quot;accuracy&quot;</span>)

<span class="hljs-keyword">def</span> <span class="hljs-title function_">compute_metrics</span>(<span class="hljs-params">eval_pred</span>):
    logits, labels = eval_pred
    predictions = np.argmax(logits, axis=-<span class="hljs-number">1</span>)
    <span class="hljs-keyword">return</span> metric.compute(predictions=predictions, references=labels)

<span class="hljs-comment"># Load pretrained model and evaluate model after each epoch</span>
model = AutoModelForSequenceClassification.from_pretrained(<span class="hljs-string">&quot;bert-base-cased&quot;</span>, num_labels=<span class="hljs-number">5</span>)
training_args = TrainingArguments(output_dir=<span class="hljs-string">&quot;test_trainer&quot;</span>, evaluation_strategy=<span class="hljs-string">&quot;epoch&quot;</span>)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=small_train_dataset,
    eval_dataset=small_eval_dataset,
    compute_metrics=compute_metrics,
)

trainer.train()`,wrap:!1}}),m=new S({props:{title:"Seq2SeqTrainer",local:"seq2seqtrainer",headingTag:"h2"}}),J=new x({props:{code:"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",highlighted:`<span class="hljs-keyword">import</span> nltk
<span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset
<span class="hljs-keyword">import</span> evaluate
<span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoTokenizer, DataCollatorForSeq2Seq
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoModelForSeq2SeqLM, Seq2SeqTrainingArguments, Seq2SeqTrainer

<span class="hljs-comment"># Prepare and tokenize dataset</span>
billsum = load_dataset(<span class="hljs-string">&quot;billsum&quot;</span>, split=<span class="hljs-string">&quot;ca_test&quot;</span>).shuffle(seed=<span class="hljs-number">42</span>).select(<span class="hljs-built_in">range</span>(<span class="hljs-number">200</span>))
billsum = billsum.train_test_split(test_size=<span class="hljs-number">0.2</span>)
tokenizer = AutoTokenizer.from_pretrained(<span class="hljs-string">&quot;t5-small&quot;</span>)
prefix = <span class="hljs-string">&quot;summarize: &quot;</span>

<span class="hljs-keyword">def</span> <span class="hljs-title function_">preprocess_function</span>(<span class="hljs-params">examples</span>):
    inputs = [prefix + doc <span class="hljs-keyword">for</span> doc <span class="hljs-keyword">in</span> examples[<span class="hljs-string">&quot;text&quot;</span>]]
    model_inputs = tokenizer(inputs, max_length=<span class="hljs-number">1024</span>, truncation=<span class="hljs-literal">True</span>)

    labels = tokenizer(text_target=examples[<span class="hljs-string">&quot;summary&quot;</span>], max_length=<span class="hljs-number">128</span>, truncation=<span class="hljs-literal">True</span>)

    model_inputs[<span class="hljs-string">&quot;labels&quot;</span>] = labels[<span class="hljs-string">&quot;input_ids&quot;</span>]
    <span class="hljs-keyword">return</span> model_inputs

tokenized_billsum = billsum.<span class="hljs-built_in">map</span>(preprocess_function, batched=<span class="hljs-literal">True</span>)

<span class="hljs-comment"># Setup evaluation</span>
nltk.download(<span class="hljs-string">&quot;punkt_tab&quot;</span>, quiet=<span class="hljs-literal">True</span>)
metric = evaluate.load(<span class="hljs-string">&quot;rouge&quot;</span>)

<span class="hljs-keyword">def</span> <span class="hljs-title function_">compute_metrics</span>(<span class="hljs-params">eval_preds</span>):
    preds, labels = eval_preds

    <span class="hljs-comment"># decode preds and labels</span>
    labels = np.where(labels != -<span class="hljs-number">100</span>, labels, tokenizer.pad_token_id)
    decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=<span class="hljs-literal">True</span>)
    decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=<span class="hljs-literal">True</span>)

    <span class="hljs-comment"># rougeLSum expects newline after each sentence</span>
    decoded_preds = [<span class="hljs-string">&quot;\\n&quot;</span>.join(nltk.sent_tokenize(pred.strip())) <span class="hljs-keyword">for</span> pred <span class="hljs-keyword">in</span> decoded_preds]
    decoded_labels = [<span class="hljs-string">&quot;\\n&quot;</span>.join(nltk.sent_tokenize(label.strip())) <span class="hljs-keyword">for</span> label <span class="hljs-keyword">in</span> decoded_labels]

    result = metric.compute(predictions=decoded_preds, references=decoded_labels, use_stemmer=<span class="hljs-literal">True</span>)
    <span class="hljs-keyword">return</span> result

<span class="hljs-comment"># Load pretrained model and evaluate model after each epoch</span>
model = AutoModelForSeq2SeqLM.from_pretrained(<span class="hljs-string">&quot;t5-small&quot;</span>)
data_collator = DataCollatorForSeq2Seq(tokenizer=tokenizer, model=model)

training_args = Seq2SeqTrainingArguments(
    output_dir=<span class="hljs-string">&quot;./results&quot;</span>,
    evaluation_strategy=<span class="hljs-string">&quot;epoch&quot;</span>,
    learning_rate=<span class="hljs-number">2e-5</span>,
    per_device_train_batch_size=<span class="hljs-number">16</span>,
    per_device_eval_batch_size=<span class="hljs-number">4</span>,
    weight_decay=<span class="hljs-number">0.01</span>,
    save_total_limit=<span class="hljs-number">3</span>,
    num_train_epochs=<span class="hljs-number">2</span>,
    fp16=<span class="hljs-literal">True</span>,
    predict_with_generate=<span class="hljs-literal">True</span>
)

trainer = Seq2SeqTrainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_billsum[<span class="hljs-string">&quot;train&quot;</span>],
    eval_dataset=tokenized_billsum[<span class="hljs-string">&quot;test&quot;</span>],
    tokenizer=tokenizer,
    data_collator=data_collator,
    compute_metrics=compute_metrics
)

trainer.train()`,wrap:!1}}),h=new ns({props:{source:"https://github.com/huggingface/evaluate/blob/main/docs/source/transformers_integrations.mdx"}}),{c(){r=T("meta"),V=t(),G=T("p"),f=t(),j(p.$$.fragment),I=t(),c=T("p"),c.textContent=Q,X=t(),j(M.$$.fragment),R=t(),j(i.$$.fragment),_=t(),o=T("p"),o.innerHTML=A,C=t(),j(d.$$.fragment),E=t(),j(m.$$.fragment),F=t(),y=T("p"),y.innerHTML=$,g=t(),j(J.$$.fragment),z=t(),u=T("p"),u.innerHTML=L,Y=t(),j(h.$$.fragment),N=t(),B=T("p"),this.h()},l(s){const l=as("svelte-u9bgzb",document.head);r=b(l,"META",{name:!0,content:!0}),l.forEach(e),V=n(s),G=b(s,"P",{}),K(G).forEach(e),f=n(s),U(p.$$.fragment,s),I=n(s),c=b(s,"P",{"data-svelte-h":!0}),v(c)!=="svelte-hhpzuu"&&(c.textContent=Q),X=n(s),U(M.$$.fragment,s),R=n(s),U(i.$$.fragment,s),_=n(s),o=b(s,"P",{"data-svelte-h":!0}),v(o)!=="svelte-d1oj50"&&(o.innerHTML=A),C=n(s),U(d.$$.fragment,s),E=n(s),U(m.$$.fragment,s),F=n(s),y=b(s,"P",{"data-svelte-h":!0}),v(y)!=="svelte-1xbwtuk"&&(y.innerHTML=$),g=n(s),U(J.$$.fragment,s),z=n(s),u=b(s,"P",{"data-svelte-h":!0}),v(u)!=="svelte-1x0y3m4"&&(u.innerHTML=L),Y=n(s),U(h.$$.fragment,s),N=n(s),B=b(s,"P",{}),K(B).forEach(e),this.h()},h(){D(r,"name","hf:doc:metadata"),D(r,"content",ps)},m(s,l){ts(document.head,r),a(s,V,l),a(s,G,l),a(s,f,l),w(p,s,l),a(s,I,l),a(s,c,l),a(s,X,l),w(M,s,l),a(s,R,l),w(i,s,l),a(s,_,l),a(s,o,l),a(s,C,l),w(d,s,l),a(s,E,l),w(m,s,l),a(s,F,l),a(s,y,l),a(s,g,l),w(J,s,l),a(s,z,l),a(s,u,l),a(s,Y,l),w(h,s,l),a(s,N,l),a(s,B,l),H=!0},p:O,i(s){H||(Z(p.$$.fragment,s),Z(M.$$.fragment,s),Z(i.$$.fragment,s),Z(d.$$.fragment,s),Z(m.$$.fragment,s),Z(J.$$.fragment,s),Z(h.$$.fragment,s),H=!0)},o(s){W(p.$$.fragment,s),W(M.$$.fragment,s),W(i.$$.fragment,s),W(d.$$.fragment,s),W(m.$$.fragment,s),W(J.$$.fragment,s),W(h.$$.fragment,s),H=!1},d(s){s&&(e(V),e(G),e(f),e(I),e(c),e(X),e(R),e(_),e(o),e(C),e(E),e(F),e(y),e(g),e(z),e(u),e(Y),e(N),e(B)),e(r),k(p,s),k(M,s),k(i,s),k(d,s),k(m,s),k(J,s),k(h,s)}}}const ps='{"title":"🤗 Transformers","local":"-transformers","sections":[{"title":"Trainer","local":"trainer","sections":[],"depth":2},{"title":"Seq2SeqTrainer","local":"seq2seqtrainer","sections":[],"depth":2}],"depth":1}';function cs(q){return ss(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class ms extends ls{constructor(r){super(),es(this,r,cs,rs,P,{})}}export{ms as component};