trl/pr_5607/en/_app/immutable/nodes/45.9f4ef431.js

import{s as At,o as Pt,n as Lt}from"../chunks/scheduler.7b731bd4.js";import{S as qt,i as Ot,e as g,s as n,c as m,h as Dt,a as w,d as l,b as s,f as Ht,g as o,j as J,k as xt,l as Kt,m as a,n as f,t as $,o as d,p as M}from"../chunks/index.cc268345.js";import{C as te,H as y,E as ee}from"../chunks/MermaidChart.svelte_svelte_type_style_lang.f0d99f98.js";import{C as j}from"../chunks/CodeBlock.169a125f.js";import{H as le,a as Vt}from"../chunks/HfOption.9f04abd1.js";function ae(U){let i,u;return i=new j({props:{code:"dHJhaW5pbmdfYXJncyUyMCUzRCUyMFNGVENvbmZpZyglMEElMjAlMjAlMjAlMjBwZXJfZGV2aWNlX3RyYWluX2JhdGNoX3NpemUlM0QxJTJDJTIwJTIwJTIzJTIwU3RhcnQlMjBzbWFsbCUwQSUyMCUyMCUyMCUyMGdyYWRpZW50X2FjY3VtdWxhdGlvbl9zdGVwcyUzRDglMkMlMjAlMjAlMjMlMjBNYWludGFpbiUyMGVmZmVjdGl2ZSUyMGJhdGNoJTIwc2l6ZSUwQSk=",highlighted:`training_args = SFTConfig(
    per_device_train_batch_size=<span class="hljs-number">1</span>,  <span class="hljs-comment"># Start small</span>
    gradient_accumulation_steps=<span class="hljs-number">8</span>,  <span class="hljs-comment"># Maintain effective batch size</span>
)`,wrap:!1}}),{c(){m(i.$$.fragment)},l(r){o(i.$$.fragment,r)},m(r,c){f(i,r,c),u=!0},p:Lt,i(r){u||($(i.$$.fragment,r),u=!0)},o(r){d(i.$$.fragment,r),u=!1},d(r){M(i,r)}}}function ne(U){let i,u;return i=new j({props:{code:"dHJhaW5pbmdfYXJncyUyMCUzRCUyMERQT0NvbmZpZyglMEElMjAlMjAlMjAlMjBwZXJfZGV2aWNlX3RyYWluX2JhdGNoX3NpemUlM0QxJTJDJTIwJTIwJTIzJTIwU3RhcnQlMjBzbWFsbCUwQSUyMCUyMCUyMCUyMGdyYWRpZW50X2FjY3VtdWxhdGlvbl9zdGVwcyUzRDglMkMlMjAlMjAlMjMlMjBNYWludGFpbiUyMGVmZmVjdGl2ZSUyMGJhdGNoJTIwc2l6ZSUwQSk=",highlighted:`training_args = DPOConfig(
    per_device_train_batch_size=<span class="hljs-number">1</span>,  <span class="hljs-comment"># Start small</span>
    gradient_accumulation_steps=<span class="hljs-number">8</span>,  <span class="hljs-comment"># Maintain effective batch size</span>
)`,wrap:!1}}),{c(){m(i.$$.fragment)},l(r){o(i.$$.fragment,r)},m(r,c){f(i,r,c),u=!0},p:Lt,i(r){u||($(i.$$.fragment,r),u=!0)},o(r){d(i.$$.fragment,r),u=!1},d(r){M(i,r)}}}function se(U){let i,u,r,c;return i=new Vt({props:{id:"batch_size",option:"SFT",$$slots:{default:[ae]},$$scope:{ctx:U}}}),r=new Vt({props:{id:"batch_size",option:"DPO",$$slots:{default:[ne]},$$scope:{ctx:U}}}),{c(){m(i.$$.fragment),u=n(),m(r.$$.fragment)},l(p){o(i.$$.fragment,p),u=s(p),o(r.$$.fragment,p)},m(p,h){f(i,p,h),a(p,u,h),f(r,p,h),c=!0},p(p,h){const T={};h&2&&(T.$$scope={dirty:h,ctx:p}),i.$set(T);const C={};h&2&&(C.$$scope={dirty:h,ctx:p}),r.$set(C)},i(p){c||($(i.$$.fragment,p),$(r.$$.fragment,p),c=!0)},o(p){d(i.$$.fragment,p),d(r.$$.fragment,p),c=!1},d(p){p&&l(u),M(i,p),M(r,p)}}}function ie(U){let i,u,r,c,p,h,T,C,R,Gt="TRL is a comprehensive library for post-training foundation models using techniques like Supervised Fine-Tuning (SFT), Group Relative Policy Optimization (GRPO), Direct Preference Optimization (DPO).",at,F,nt,_,Xt="Get started instantly with TRL’s most popular trainers. Each example uses compact models for quick experimentation.",st,Z,it,I,rt,k,pt,W,mt,G,ot,X,ft,B,$t,E,dt,Q,Mt,z,Bt="Skip the code entirely - train directly from your terminal:",ut,S,yt,Y,ct,v,ht,N,Et='<li><a href="sft_trainer">SFT Trainer</a> - Complete SFT guide</li> <li><a href="dpo_trainer">DPO Trainer</a> - Preference alignment</li> <li><a href="grpo_trainer">GRPO Trainer</a> - Group relative policy optimization</li>',gt,H,wt,x,Qt='<li><a href="distributing_training">Distributed Training</a> - Multi-GPU setups</li> <li><a href="reducing_memory_usage">Memory Optimization</a> - Efficient training</li> <li><a href="peft_integration">PEFT Integration</a> - LoRA and QLoRA</li>',Tt,V,Jt,L,zt='<li><a href="https://github.com/huggingface/trl/tree/main/examples" rel="nofollow">Example Scripts</a> - Production-ready code</li> <li><a href="community_tutorials">Community Tutorials</a> - External guides</li>',Ut,A,bt,P,jt,q,St="Reduce batch size and enable optimizations:",Ct,b,Rt,O,Ft,D,Yt="Try adjusting the learning rate:",_t,K,Zt,tt,vt='For more help, open an <a href="https://github.com/huggingface/trl/issues" rel="nofollow">issue on GitHub</a>.',It,et,kt,lt,Wt;return p=new te({props:{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"}}),T=new y({props:{title:"Quickstart",local:"quickstart",headingTag:"h1"}}),F=new y({props:{title:"Quick Examples",local:"quick-examples",headingTag:"h2"}}),Z=new y({props:{title:"Supervised Fine-Tuning",local:"supervised-fine-tuning",headingTag:"h3"}}),I=new j({props:{code:"ZnJvbSUyMHRybCUyMGltcG9ydCUyMFNGVFRyYWluZXIlMEFmcm9tJTIwZGF0YXNldHMlMjBpbXBvcnQlMjBsb2FkX2RhdGFzZXQlMEElMEF0cmFpbmVyJTIwJTNEJTIwU0ZUVHJhaW5lciglMEElMjAlMjAlMjAlMjBtb2RlbCUzRCUyMlF3ZW4lMkZRd2VuMi41LTAuNUIlMjIlMkMlMEElMjAlMjAlMjAlMjB0cmFpbl9kYXRhc2V0JTNEbG9hZF9kYXRhc2V0KCUyMnRybC1saWIlMkZDYXB5YmFyYSUyMiUyQyUyMHNwbGl0JTNEJTIydHJhaW4lMjIpJTJDJTBBKSUwQXRyYWluZXIudHJhaW4oKQ==",highlighted:`<span class="hljs-keyword">from</span> trl <span class="hljs-keyword">import</span> SFTTrainer
<span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset

trainer = SFTTrainer(
    model=<span class="hljs-string">&quot;Qwen/Qwen2.5-0.5B&quot;</span>,
    train_dataset=load_dataset(<span class="hljs-string">&quot;trl-lib/Capybara&quot;</span>, split=<span class="hljs-string">&quot;train&quot;</span>),
)
trainer.train()`,wrap:!1}}),k=new y({props:{title:"Group Relative Policy Optimization",local:"group-relative-policy-optimization",headingTag:"h3"}}),W=new j({props:{code:"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",highlighted:`<span class="hljs-keyword">from</span> trl <span class="hljs-keyword">import</span> GRPOTrainer
<span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset
<span class="hljs-keyword">from</span> trl.rewards <span class="hljs-keyword">import</span> accuracy_reward

trainer = GRPOTrainer(
    model=<span class="hljs-string">&quot;Qwen/Qwen2.5-0.5B-Instruct&quot;</span>,
    train_dataset=load_dataset(<span class="hljs-string">&quot;trl-lib/DeepMath-103K&quot;</span>, split=<span class="hljs-string">&quot;train&quot;</span>),
    reward_funcs=accuracy_reward,
)
trainer.train()`,wrap:!1}}),G=new y({props:{title:"Direct Preference Optimization",local:"direct-preference-optimization",headingTag:"h3"}}),X=new j({props:{code:"ZnJvbSUyMHRybCUyMGltcG9ydCUyMERQT1RyYWluZXIlMEFmcm9tJTIwZGF0YXNldHMlMjBpbXBvcnQlMjBsb2FkX2RhdGFzZXQlMEElMEF0cmFpbmVyJTIwJTNEJTIwRFBPVHJhaW5lciglMEElMjAlMjAlMjAlMjBtb2RlbCUzRCUyMlF3ZW4lMkZRd2VuMi41LTAuNUItSW5zdHJ1Y3QlMjIlMkMlMEElMjAlMjAlMjAlMjB0cmFpbl9kYXRhc2V0JTNEbG9hZF9kYXRhc2V0KCUyMnRybC1saWIlMkZ1bHRyYWZlZWRiYWNrX2JpbmFyaXplZCUyMiUyQyUyMHNwbGl0JTNEJTIydHJhaW4lMjIpJTJDJTBBKSUwQXRyYWluZXIudHJhaW4oKQ==",highlighted:`<span class="hljs-keyword">from</span> trl <span class="hljs-keyword">import</span> DPOTrainer
<span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset

trainer = DPOTrainer(
    model=<span class="hljs-string">&quot;Qwen/Qwen2.5-0.5B-Instruct&quot;</span>,
    train_dataset=load_dataset(<span class="hljs-string">&quot;trl-lib/ultrafeedback_binarized&quot;</span>, split=<span class="hljs-string">&quot;train&quot;</span>),
)
trainer.train()`,wrap:!1}}),B=new y({props:{title:"Reward Modeling",local:"reward-modeling",headingTag:"h3"}}),E=new j({props:{code:"ZnJvbSUyMHRybCUyMGltcG9ydCUyMFJld2FyZFRyYWluZXIlMEFmcm9tJTIwZGF0YXNldHMlMjBpbXBvcnQlMjBsb2FkX2RhdGFzZXQlMEElMEFkYXRhc2V0JTIwJTNEJTIwbG9hZF9kYXRhc2V0KCUyMnRybC1saWIlMkZ1bHRyYWZlZWRiYWNrX2JpbmFyaXplZCUyMiUyQyUyMHNwbGl0JTNEJTIydHJhaW4lMjIpJTBBJTBBdHJhaW5lciUyMCUzRCUyMFJld2FyZFRyYWluZXIoJTBBJTIwJTIwJTIwJTIwbW9kZWwlM0QlMjJRd2VuJTJGUXdlbjIuNS0wLjVCLUluc3RydWN0JTIyJTJDJTBBJTIwJTIwJTIwJTIwdHJhaW5fZGF0YXNldCUzRGRhdGFzZXQlMkMlMEEpJTBBdHJhaW5lci50cmFpbigp",highlighted:`<span class="hljs-keyword">from</span> trl <span class="hljs-keyword">import</span> RewardTrainer
<span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset

dataset = load_dataset(<span class="hljs-string">&quot;trl-lib/ultrafeedback_binarized&quot;</span>, split=<span class="hljs-string">&quot;train&quot;</span>)

trainer = RewardTrainer(
    model=<span class="hljs-string">&quot;Qwen/Qwen2.5-0.5B-Instruct&quot;</span>,
    train_dataset=dataset,
)
trainer.train()`,wrap:!1}}),Q=new y({props:{title:"Command Line Interface",local:"command-line-interface",headingTag:"h2"}}),S=new j({props:{code:"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",highlighted:`<span class="hljs-comment"># SFT: Fine-tune on instructions</span>
trl sft --model_name_or_path Qwen/Qwen2.5-0.5B \\
    --dataset_name trl-lib/Capybara

<span class="hljs-comment"># DPO: Align with preferences  </span>
trl dpo --model_name_or_path Qwen/Qwen2.5-0.5B-Instruct \\
    --dataset_name trl-lib/ultrafeedback_binarized

<span class="hljs-comment"># Reward: Train a reward model</span>
trl reward --model_name_or_path Qwen/Qwen2.5-0.5B-Instruct \\
    --dataset_name trl-lib/ultrafeedback_binarized`,wrap:!1}}),Y=new y({props:{title:"What’s Next?",local:"whats-next",headingTag:"h2"}}),v=new y({props:{title:"📚 Learn More",local:"-learn-more",headingTag:"h3"}}),H=new y({props:{title:"🚀 Scale Up",local:"-scale-up",headingTag:"h3"}}),V=new y({props:{title:"💡 Examples",local:"-examples",headingTag:"h3"}}),A=new y({props:{title:"Troubleshooting",local:"troubleshooting",headingTag:"h2"}}),P=new y({props:{title:"Out of Memory?",local:"out-of-memory",headingTag:"h3"}}),b=new le({props:{id:"batch_size",options:["SFT","DPO"],$$slots:{default:[se]},$$scope:{ctx:U}}}),O=new y({props:{title:"Loss not decreasing?",local:"loss-not-decreasing",headingTag:"h3"}}),K=new j({props:{code:"dHJhaW5pbmdfYXJncyUyMCUzRCUyMFNGVENvbmZpZyhsZWFybmluZ19yYXRlJTNEMmUtNSklMjAlMjAlMjMlMjBHb29kJTIwc3RhcnRpbmclMjBwb2ludA==",highlighted:'training_args = SFTConfig(learning_rate=<span class="hljs-number">2e-5</span>)  <span class="hljs-comment"># Good starting point</span>',wrap:!1}}),et=new ee({props:{source:"https://github.com/huggingface/trl/blob/main/docs/source/quickstart.md"}}),{c(){i=g("meta"),u=n(),r=g("p"),c=n(),m(p.$$.fragment),h=n(),m(T.$$.fragment),C=n(),R=g("p"),R.textContent=Gt,at=n(),m(F.$$.fragment),nt=n(),_=g("p"),_.textContent=Xt,st=n(),m(Z.$$.fragment),it=n(),m(I.$$.fragment),rt=n(),m(k.$$.fragment),pt=n(),m(W.$$.fragment),mt=n(),m(G.$$.fragment),ot=n(),m(X.$$.fragment),ft=n(),m(B.$$.fragment),$t=n(),m(E.$$.fragment),dt=n(),m(Q.$$.fragment),Mt=n(),z=g("p"),z.textContent=Bt,ut=n(),m(S.$$.fragment),yt=n(),m(Y.$$.fragment),ct=n(),m(v.$$.fragment),ht=n(),N=g("ul"),N.innerHTML=Et,gt=n(),m(H.$$.fragment),wt=n(),x=g("ul"),x.innerHTML=Qt,Tt=n(),m(V.$$.fragment),Jt=n(),L=g("ul"),L.innerHTML=zt,Ut=n(),m(A.$$.fragment),bt=n(),m(P.$$.fragment),jt=n(),q=g("p"),q.textContent=St,Ct=n(),m(b.$$.fragment),Rt=n(),m(O.$$.fragment),Ft=n(),D=g("p"),D.textContent=Yt,_t=n(),m(K.$$.fragment),Zt=n(),tt=g("p"),tt.innerHTML=vt,It=n(),m(et.$$.fragment),kt=n(),lt=g("p"),this.h()},l(t){const e=Dt("svelte-u9bgzb",document.head);i=w(e,"META",{name:!0,content:!0}),e.forEach(l),u=s(t),r=w(t,"P",{}),Ht(r).forEach(l),c=s(t),o(p.$$.fragment,t),h=s(t),o(T.$$.fragment,t),C=s(t),R=w(t,"P",{"data-svelte-h":!0}),J(R)!=="svelte-mfh7lp"&&(R.textContent=Gt),at=s(t),o(F.$$.fragment,t),nt=s(t),_=w(t,"P",{"data-svelte-h":!0}),J(_)!=="svelte-1l0bo87"&&(_.textContent=Xt),st=s(t),o(Z.$$.fragment,t),it=s(t),o(I.$$.fragment,t),rt=s(t),o(k.$$.fragment,t),pt=s(t),o(W.$$.fragment,t),mt=s(t),o(G.$$.fragment,t),ot=s(t),o(X.$$.fragment,t),ft=s(t),o(B.$$.fragment,t),$t=s(t),o(E.$$.fragment,t),dt=s(t),o(Q.$$.fragment,t),Mt=s(t),z=w(t,"P",{"data-svelte-h":!0}),J(z)!=="svelte-kedgej"&&(z.textContent=Bt),ut=s(t),o(S.$$.fragment,t),yt=s(t),o(Y.$$.fragment,t),ct=s(t),o(v.$$.fragment,t),ht=s(t),N=w(t,"UL",{"data-svelte-h":!0}),J(N)!=="svelte-8iv07w"&&(N.innerHTML=Et),gt=s(t),o(H.$$.fragment,t),wt=s(t),x=w(t,"UL",{"data-svelte-h":!0}),J(x)!=="svelte-li8j2i"&&(x.innerHTML=Qt),Tt=s(t),o(V.$$.fragment,t),Jt=s(t),L=w(t,"UL",{"data-svelte-h":!0}),J(L)!=="svelte-qpvndh"&&(L.innerHTML=zt),Ut=s(t),o(A.$$.fragment,t),bt=s(t),o(P.$$.fragment,t),jt=s(t),q=w(t,"P",{"data-svelte-h":!0}),J(q)!=="svelte-6o3jij"&&(q.textContent=St),Ct=s(t),o(b.$$.fragment,t),Rt=s(t),o(O.$$.fragment,t),Ft=s(t),D=w(t,"P",{"data-svelte-h":!0}),J(D)!=="svelte-1o1e4dh"&&(D.textContent=Yt),_t=s(t),o(K.$$.fragment,t),Zt=s(t),tt=w(t,"P",{"data-svelte-h":!0}),J(tt)!=="svelte-4qh17o"&&(tt.innerHTML=vt),It=s(t),o(et.$$.fragment,t),kt=s(t),lt=w(t,"P",{}),Ht(lt).forEach(l),this.h()},h(){xt(i,"name","hf:doc:metadata"),xt(i,"content",re)},m(t,e){Kt(document.head,i),a(t,u,e),a(t,r,e),a(t,c,e),f(p,t,e),a(t,h,e),f(T,t,e),a(t,C,e),a(t,R,e),a(t,at,e),f(F,t,e),a(t,nt,e),a(t,_,e),a(t,st,e),f(Z,t,e),a(t,it,e),f(I,t,e),a(t,rt,e),f(k,t,e),a(t,pt,e),f(W,t,e),a(t,mt,e),f(G,t,e),a(t,ot,e),f(X,t,e),a(t,ft,e),f(B,t,e),a(t,$t,e),f(E,t,e),a(t,dt,e),f(Q,t,e),a(t,Mt,e),a(t,z,e),a(t,ut,e),f(S,t,e),a(t,yt,e),f(Y,t,e),a(t,ct,e),f(v,t,e),a(t,ht,e),a(t,N,e),a(t,gt,e),f(H,t,e),a(t,wt,e),a(t,x,e),a(t,Tt,e),f(V,t,e),a(t,Jt,e),a(t,L,e),a(t,Ut,e),f(A,t,e),a(t,bt,e),f(P,t,e),a(t,jt,e),a(t,q,e),a(t,Ct,e),f(b,t,e),a(t,Rt,e),f(O,t,e),a(t,Ft,e),a(t,D,e),a(t,_t,e),f(K,t,e),a(t,Zt,e),a(t,tt,e),a(t,It,e),f(et,t,e),a(t,kt,e),a(t,lt,e),Wt=!0},p(t,[e]){const Nt={};e&2&&(Nt.$$scope={dirty:e,ctx:t}),b.$set(Nt)},i(t){Wt||($(p.$$.fragment,t),$(T.$$.fragment,t),$(F.$$.fragment,t),$(Z.$$.fragment,t),$(I.$$.fragment,t),$(k.$$.fragment,t),$(W.$$.fragment,t),$(G.$$.fragment,t),$(X.$$.fragment,t),$(B.$$.fragment,t),$(E.$$.fragment,t),$(Q.$$.fragment,t),$(S.$$.fragment,t),$(Y.$$.fragment,t),$(v.$$.fragment,t),$(H.$$.fragment,t),$(V.$$.fragment,t),$(A.$$.fragment,t),$(P.$$.fragment,t),$(b.$$.fragment,t),$(O.$$.fragment,t),$(K.$$.fragment,t),$(et.$$.fragment,t),Wt=!0)},o(t){d(p.$$.fragment,t),d(T.$$.fragment,t),d(F.$$.fragment,t),d(Z.$$.fragment,t),d(I.$$.fragment,t),d(k.$$.fragment,t),d(W.$$.fragment,t),d(G.$$.fragment,t),d(X.$$.fragment,t),d(B.$$.fragment,t),d(E.$$.fragment,t),d(Q.$$.fragment,t),d(S.$$.fragment,t),d(Y.$$.fragment,t),d(v.$$.fragment,t),d(H.$$.fragment,t),d(V.$$.fragment,t),d(A.$$.fragment,t),d(P.$$.fragment,t),d(b.$$.fragment,t),d(O.$$.fragment,t),d(K.$$.fragment,t),d(et.$$.fragment,t),Wt=!1},d(t){t&&(l(u),l(r),l(c),l(h),l(C),l(R),l(at),l(nt),l(_),l(st),l(it),l(rt),l(pt),l(mt),l(ot),l(ft),l($t),l(dt),l(Mt),l(z),l(ut),l(yt),l(ct),l(ht),l(N),l(gt),l(wt),l(x),l(Tt),l(Jt),l(L),l(Ut),l(bt),l(jt),l(q),l(Ct),l(Rt),l(Ft),l(D),l(_t),l(Zt),l(tt),l(It),l(kt),l(lt)),l(i),M(p,t),M(T,t),M(F,t),M(Z,t),M(I,t),M(k,t),M(W,t),M(G,t),M(X,t),M(B,t),M(E,t),M(Q,t),M(S,t),M(Y,t),M(v,t),M(H,t),M(V,t),M(A,t),M(P,t),M(b,t),M(O,t),M(K,t),M(et,t)}}}const re='{"title":"Quickstart","local":"quickstart","sections":[{"title":"Quick Examples","local":"quick-examples","sections":[{"title":"Supervised Fine-Tuning","local":"supervised-fine-tuning","sections":[],"depth":3},{"title":"Group Relative Policy Optimization","local":"group-relative-policy-optimization","sections":[],"depth":3},{"title":"Direct Preference Optimization","local":"direct-preference-optimization","sections":[],"depth":3},{"title":"Reward Modeling","local":"reward-modeling","sections":[],"depth":3}],"depth":2},{"title":"Command Line Interface","local":"command-line-interface","sections":[],"depth":2},{"title":"What’s Next?","local":"whats-next","sections":[{"title":"📚 Learn More","local":"-learn-more","sections":[],"depth":3},{"title":"🚀 Scale Up","local":"-scale-up","sections":[],"depth":3},{"title":"💡 Examples","local":"-examples","sections":[],"depth":3}],"depth":2},{"title":"Troubleshooting","local":"troubleshooting","sections":[{"title":"Out of Memory?","local":"out-of-memory","sections":[],"depth":3},{"title":"Loss not decreasing?","local":"loss-not-decreasing","sections":[],"depth":3}],"depth":2}],"depth":1}';function pe(U){return Pt(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class Me extends qt{constructor(i){super(),Ot(this,i,pe,ie,At,{})}}export{Me as component};