bitsandbytes/pr_1501/en/_app/immutable/nodes/6.05e4e321.js

import{s as $e,n as Be,o as Ee}from"../chunks/scheduler.852ec091.js";import{S as Ae,i as Re,g as i,s as a,r,A as Ze,h as M,f as s,c as n,j as Qe,u as p,x as c,k as ge,y as We,a as t,v as m,d as o,t as y,w}from"../chunks/index.28275fd3.js";import{C as X}from"../chunks/CodeBlock.c3366071.js";import{H as N,E as ve}from"../chunks/EditOnGithub.582011f0.js";function ze(Te){let T,G,L,F,J,H,U,Je="The below academic work is ordered in reverse chronological order.",Y,u,D,j,Ue="Authors: Tim Dettmers, Ruslan Svirschevski, Vage Egiazarian, Denis Kuznedelev, Elias Frantar, Saleh Ashkboos, Alexander Borzunov, Torsten Hoefler, Dan Alistarh",V,h,ue='<li><a href="https://twitter.com/Tim_Dettmers/status/1666076553665744896" rel="nofollow">Twitter summary thread</a></li>',_,d,P,b,q,f,je="Authors: Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, Luke Zettlemoyer",O,C,he='<li><a href="https://www.youtube.com/watch?v=y9PHWGOa8HA&amp;ab_channel=LondonMachineLearningMeetup" rel="nofollow">Video</a></li> <li><a href="https://twitter.com/Tim_Dettmers/status/1661379354507476994" rel="nofollow">Twitter summary thread</a></li>',K,I,ee,k,le,Q,de="Authors: Tim Dettmers, Luke Zettlemoyer",se,g,be='<li><a href="https://www.youtube.com/watch?v=odlQa6AE1gY&amp;ab_channel=TheInsideView" rel="nofollow">Video</a></li> <li><a href="https://twitter.com/Tim_Dettmers/status/1605209171758284805" rel="nofollow">Twitter summary thread</a></li>',te,$,ae,B,ne,E,fe="Authors: Tim Dettmers, Mike Lewis, Younes Belkada, Luke Zettlemoyer",ie,A,Ce='<li><a href="https://huggingface.co/blog/hf-bitsandbytes-integration" rel="nofollow">LLM.int8() Blog Post</a></li> <li><a href="https://timdettmers.com/2022/08/17/llm-int8-and-emergent-features/" rel="nofollow">LLM.int8() Emergent Features Blog Post</a></li> <li><a href="https://towardsdatascience.com/introduction-to-weight-quantization-2494701b9c0c" rel="nofollow">Introduction to Weight Quantization</a></li> <li><a href="https://twitter.com/Tim_Dettmers/status/1598351301942951937" rel="nofollow">Poster</a></li>',Me,R,re,Z,pe,W,Ie="Authors: Tim Dettmers, Mike Lewis, Sam Shleifer, Luke Zettlemoyer",me,v,ke='<li><a href="https://www.youtube.com/watch?v=IxrlHAJtqKE" rel="nofollow">Video</a></li> <li><a href="https://twitter.com/Tim_Dettmers/status/1446472128979562499" rel="nofollow">Twitter summary thread</a></li>',oe,z,ye,x,we,S,ce;return J=new N({props:{title:"Papers, related resources & how to cite",local:"papers-related-resources--how-to-cite",headingTag:"h1"}}),u=new N({props:{title:"SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression (Jun 2023)",local:"spqr-a-sparse-quantized-representation-for-near-lossless-llm-weight-compression-jun-2023",headingTag:"h2"}}),d=new X({props:{code:"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",highlighted:`@article{dettmers2023spqr,
  title={SpQR: A Sparse-Quantized Representation for Near-Lossless <span class="hljs-keyword">LLM </span>Weight Compression},
  author={Dettmers, Tim <span class="hljs-keyword">and </span>Svirschevski, Ruslan <span class="hljs-keyword">and </span>Egiazarian, Vage <span class="hljs-keyword">and </span>Kuznedelev, Denis <span class="hljs-keyword">and </span>Frantar, Elias <span class="hljs-keyword">and </span>Ashkboos, Saleh <span class="hljs-keyword">and </span><span class="hljs-keyword">Borzunov, </span>Alexander <span class="hljs-keyword">and </span>Hoefler, Torsten <span class="hljs-keyword">and </span>Alistarh, Dan},
  <span class="hljs-keyword">journal={arXiv </span>preprint arXiv:<span class="hljs-number">2306</span>.<span class="hljs-number">03078</span>},
  year={<span class="hljs-number">2023</span>}
}`,wrap:!1}}),b=new N({props:{title:"QLoRA: Efficient Finetuning of Quantized LLMs (May 2023)",local:"qlora-efficient-finetuning-of-quantized-llms-may-2023",headingTag:"h2"}}),I=new X({props:{code:"JTQwYXJ0aWNsZSU3QmRldHRtZXJzMjAyM3Fsb3JhJTJDJTBBJTIwJTIwdGl0bGUlM0QlN0JRbG9yYSUzQSUyMEVmZmljaWVudCUyMGZpbmV0dW5pbmclMjBvZiUyMHF1YW50aXplZCUyMGxsbXMlN0QlMkMlMEElMjAlMjBhdXRob3IlM0QlN0JEZXR0bWVycyUyQyUyMFRpbSUyMGFuZCUyMFBhZ25vbmklMkMlMjBBcnRpZG9ybyUyMGFuZCUyMEhvbHR6bWFuJTJDJTIwQXJpJTIwYW5kJTIwWmV0dGxlbW95ZXIlMkMlMjBMdWtlJTdEJTJDJTBBJTIwJTIwam91cm5hbCUzRCU3QmFyWGl2JTIwcHJlcHJpbnQlMjBhclhpdiUzQTIzMDUuMTQzMTQlN0QlMkMlMEElMjAlMjB5ZWFyJTNEJTdCMjAyMyU3RCUwQSU3RA==",highlighted:`@article{dettmers2023qlora,
  title={Qlora: Efficient finetuning of quantized <span class="hljs-keyword">llms},
</span>  author={Dettmers, Tim <span class="hljs-keyword">and </span>Pagnoni, Artidoro <span class="hljs-keyword">and </span>Holtzman, Ari <span class="hljs-keyword">and </span>Zettlemoyer, Luke},
  <span class="hljs-keyword">journal={arXiv </span>preprint arXiv:<span class="hljs-number">2305</span>.<span class="hljs-number">14314</span>},
  year={<span class="hljs-number">2023</span>}
}`,wrap:!1}}),k=new N({props:{title:"The case for 4-bit precision: k-bit Inference Scaling Laws (Dec 2022)",local:"the-case-for-4-bit-precision-k-bit-inference-scaling-laws-dec-2022",headingTag:"h2"}}),$=new X({props:{code:"JTQwaW5wcm9jZWVkaW5ncyU3QmRldHRtZXJzMjAyM2Nhc2UlMkMlMEElMjAlMjB0aXRsZSUzRCU3QlRoZSUyMGNhc2UlMjBmb3IlMjA0LWJpdCUyMHByZWNpc2lvbiUzQSUyMGstYml0JTIwaW5mZXJlbmNlJTIwc2NhbGluZyUyMGxhd3MlN0QlMkMlMEElMjAlMjBhdXRob3IlM0QlN0JEZXR0bWVycyUyQyUyMFRpbSUyMGFuZCUyMFpldHRsZW1veWVyJTJDJTIwTHVrZSU3RCUyQyUwQSUyMCUyMGJvb2t0aXRsZSUzRCU3QkludGVybmF0aW9uYWwlMjBDb25mZXJlbmNlJTIwb24lMjBNYWNoaW5lJTIwTGVhcm5pbmclN0QlMkMlMEElMjAlMjBwYWdlcyUzRCU3Qjc3NTAtLTc3NzQlN0QlMkMlMEElMjAlMjB5ZWFyJTNEJTdCMjAyMyU3RCUyQyUwQSUyMCUyMG9yZ2FuaXphdGlvbiUzRCU3QlBNTFIlN0QlMEElN0Q=",highlighted:`<span class="language-xml">@inproceedings</span><span class="hljs-template-variable">{dettmers2023case,
  title={The case for 4-bit precision: k-bit inference scaling laws}</span><span class="language-xml">,
  author=</span><span class="hljs-template-variable">{Dettmers, Tim and Zettlemoyer, Luke}</span><span class="language-xml">,
  booktitle=</span><span class="hljs-template-variable">{International Conference on Machine Learning}</span><span class="language-xml">,
  pages=</span><span class="hljs-template-variable">{7750--7774}</span><span class="language-xml">,
  year=</span><span class="hljs-template-variable">{2023}</span><span class="language-xml">,
  organization=</span><span class="hljs-template-variable">{PMLR}</span><span class="language-xml">
}</span>`,wrap:!1}}),B=new N({props:{title:"LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale (Nov 2022)",local:"llm-int8",headingTag:"h2"}}),R=new X({props:{code:"JTQwYXJ0aWNsZSU3QmRldHRtZXJzMjAyMmxsbSUyQyUwQSUyMCUyMHRpdGxlJTNEJTdCTGxtLiUyMGludDglMjAoKSUzQSUyMDgtYml0JTIwbWF0cml4JTIwbXVsdGlwbGljYXRpb24lMjBmb3IlMjB0cmFuc2Zvcm1lcnMlMjBhdCUyMHNjYWxlJTdEJTJDJTBBJTIwJTIwYXV0aG9yJTNEJTdCRGV0dG1lcnMlMkMlMjBUaW0lMjBhbmQlMjBMZXdpcyUyQyUyME1pa2UlMjBhbmQlMjBCZWxrYWRhJTJDJTIwWW91bmVzJTIwYW5kJTIwWmV0dGxlbW95ZXIlMkMlMjBMdWtlJTdEJTJDJTBBJTIwJTIwam91cm5hbCUzRCU3QmFyWGl2JTIwcHJlcHJpbnQlMjBhclhpdiUzQTIyMDguMDczMzklN0QlMkMlMEElMjAlMjB5ZWFyJTNEJTdCMjAyMiU3RCUwQSU3RA==",highlighted:`@article{dettmers2022llm,
  title={<span class="hljs-keyword">Llm. </span>int8 (): <span class="hljs-number">8</span>-<span class="hljs-keyword">bit </span>matrix <span class="hljs-keyword">multiplication </span>for transformers <span class="hljs-built_in">at</span> <span class="hljs-keyword">scale},
</span>  author={Dettmers, Tim <span class="hljs-keyword">and </span>Lewis, Mike <span class="hljs-keyword">and </span><span class="hljs-keyword">Belkada, </span>Younes <span class="hljs-keyword">and </span>Zettlemoyer, Luke},
  <span class="hljs-keyword">journal={arXiv </span>preprint arXiv:<span class="hljs-number">2208</span>.<span class="hljs-number">07339</span>},
  year={<span class="hljs-number">2022</span>}
}`,wrap:!1}}),Z=new N({props:{title:"8-bit Optimizers via Block-wise Quantization (Oct 2021)",local:"8-bit-optimizers-via-block-wise-quantization-oct-2021",headingTag:"h2"}}),z=new X({props:{code:"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",highlighted:`@article{DBLP:journals<span class="hljs-regexp">/corr/</span>abs-<span class="hljs-number">2110</span>-<span class="hljs-number">02861</span>,
  author       = {Tim Dettmers and
                  Mike Lewis and
                  Sam Shleifer and
                  Luke Zettlemoyer},
  title        = {<span class="hljs-number">8</span>-bit Optimizers via Block-wise Quantization},
  journal      = {CoRR},
  volume       = {abs/<span class="hljs-number">2110.02861</span>},
  year         = {<span class="hljs-number">2021</span>},
  url          = {https:<span class="hljs-regexp">//</span>arxiv.org<span class="hljs-regexp">/abs/</span><span class="hljs-number">2110.02861</span>},
  eprinttype    = {arXiv},
  eprint       = {<span class="hljs-number">2110.02861</span>},
  timestamp    = {Thu, <span class="hljs-number">21</span> Oct <span class="hljs-number">2021</span> <span class="hljs-number">16</span>:<span class="hljs-number">20</span>:<span class="hljs-number">08</span> +<span class="hljs-number">0200</span>},
  biburl       = {https:<span class="hljs-regexp">//</span>dblp.org<span class="hljs-regexp">/rec/</span>journals<span class="hljs-regexp">/corr/</span>abs-<span class="hljs-number">2110</span>-<span class="hljs-number">02861</span>.bib},
  bibsource    = {dblp computer science bibliography, https:<span class="hljs-regexp">//</span>dblp.org}
}`,wrap:!1}}),x=new ve({props:{source:"https://github.com/bitsandbytes-foundation/bitsandbytes/blob/main/docs/source/explanations/resources.mdx"}}),{c(){T=i("meta"),G=a(),L=i("p"),F=a(),r(J.$$.fragment),H=a(),U=i("p"),U.textContent=Je,Y=a(),r(u.$$.fragment),D=a(),j=i("p"),j.textContent=Ue,V=a(),h=i("ul"),h.innerHTML=ue,_=a(),r(d.$$.fragment),P=a(),r(b.$$.fragment),q=a(),f=i("p"),f.textContent=je,O=a(),C=i("ul"),C.innerHTML=he,K=a(),r(I.$$.fragment),ee=a(),r(k.$$.fragment),le=a(),Q=i("p"),Q.textContent=de,se=a(),g=i("ul"),g.innerHTML=be,te=a(),r($.$$.fragment),ae=a(),r(B.$$.fragment),ne=a(),E=i("p"),E.textContent=fe,ie=a(),A=i("ul"),A.innerHTML=Ce,Me=a(),r(R.$$.fragment),re=a(),r(Z.$$.fragment),pe=a(),W=i("p"),W.textContent=Ie,me=a(),v=i("ul"),v.innerHTML=ke,oe=a(),r(z.$$.fragment),ye=a(),r(x.$$.fragment),we=a(),S=i("p"),this.h()},l(e){const l=Ze("svelte-u9bgzb",document.head);T=M(l,"META",{name:!0,content:!0}),l.forEach(s),G=n(e),L=M(e,"P",{}),Qe(L).forEach(s),F=n(e),p(J.$$.fragment,e),H=n(e),U=M(e,"P",{"data-svelte-h":!0}),c(U)!=="svelte-1jp7sfc"&&(U.textContent=Je),Y=n(e),p(u.$$.fragment,e),D=n(e),j=M(e,"P",{"data-svelte-h":!0}),c(j)!=="svelte-6ej7zg"&&(j.textContent=Ue),V=n(e),h=M(e,"UL",{"data-svelte-h":!0}),c(h)!=="svelte-1bk8yei"&&(h.innerHTML=ue),_=n(e),p(d.$$.fragment,e),P=n(e),p(b.$$.fragment,e),q=n(e),f=M(e,"P",{"data-svelte-h":!0}),c(f)!=="svelte-1m3ezda"&&(f.textContent=je),O=n(e),C=M(e,"UL",{"data-svelte-h":!0}),c(C)!=="svelte-95bqwh"&&(C.innerHTML=he),K=n(e),p(I.$$.fragment,e),ee=n(e),p(k.$$.fragment,e),le=n(e),Q=M(e,"P",{"data-svelte-h":!0}),c(Q)!=="svelte-1ltttkj"&&(Q.textContent=de),se=n(e),g=M(e,"UL",{"data-svelte-h":!0}),c(g)!=="svelte-1uuma5m"&&(g.innerHTML=be),te=n(e),p($.$$.fragment,e),ae=n(e),p(B.$$.fragment,e),ne=n(e),E=M(e,"P",{"data-svelte-h":!0}),c(E)!=="svelte-1d05yau"&&(E.textContent=fe),ie=n(e),A=M(e,"UL",{"data-svelte-h":!0}),c(A)!=="svelte-1w2oslr"&&(A.innerHTML=Ce),Me=n(e),p(R.$$.fragment,e),re=n(e),p(Z.$$.fragment,e),pe=n(e),W=M(e,"P",{"data-svelte-h":!0}),c(W)!=="svelte-cvnv0k"&&(W.textContent=Ie),me=n(e),v=M(e,"UL",{"data-svelte-h":!0}),c(v)!=="svelte-9u33gj"&&(v.innerHTML=ke),oe=n(e),p(z.$$.fragment,e),ye=n(e),p(x.$$.fragment,e),we=n(e),S=M(e,"P",{}),Qe(S).forEach(s),this.h()},h(){ge(T,"name","hf:doc:metadata"),ge(T,"content",xe)},m(e,l){We(document.head,T),t(e,G,l),t(e,L,l),t(e,F,l),m(J,e,l),t(e,H,l),t(e,U,l),t(e,Y,l),m(u,e,l),t(e,D,l),t(e,j,l),t(e,V,l),t(e,h,l),t(e,_,l),m(d,e,l),t(e,P,l),m(b,e,l),t(e,q,l),t(e,f,l),t(e,O,l),t(e,C,l),t(e,K,l),m(I,e,l),t(e,ee,l),m(k,e,l),t(e,le,l),t(e,Q,l),t(e,se,l),t(e,g,l),t(e,te,l),m($,e,l),t(e,ae,l),m(B,e,l),t(e,ne,l),t(e,E,l),t(e,ie,l),t(e,A,l),t(e,Me,l),m(R,e,l),t(e,re,l),m(Z,e,l),t(e,pe,l),t(e,W,l),t(e,me,l),t(e,v,l),t(e,oe,l),m(z,e,l),t(e,ye,l),m(x,e,l),t(e,we,l),t(e,S,l),ce=!0},p:Be,i(e){ce||(o(J.$$.fragment,e),o(u.$$.fragment,e),o(d.$$.fragment,e),o(b.$$.fragment,e),o(I.$$.fragment,e),o(k.$$.fragment,e),o($.$$.fragment,e),o(B.$$.fragment,e),o(R.$$.fragment,e),o(Z.$$.fragment,e),o(z.$$.fragment,e),o(x.$$.fragment,e),ce=!0)},o(e){y(J.$$.fragment,e),y(u.$$.fragment,e),y(d.$$.fragment,e),y(b.$$.fragment,e),y(I.$$.fragment,e),y(k.$$.fragment,e),y($.$$.fragment,e),y(B.$$.fragment,e),y(R.$$.fragment,e),y(Z.$$.fragment,e),y(z.$$.fragment,e),y(x.$$.fragment,e),ce=!1},d(e){e&&(s(G),s(L),s(F),s(H),s(U),s(Y),s(D),s(j),s(V),s(h),s(_),s(P),s(q),s(f),s(O),s(C),s(K),s(ee),s(le),s(Q),s(se),s(g),s(te),s(ae),s(ne),s(E),s(ie),s(A),s(Me),s(re),s(pe),s(W),s(me),s(v),s(oe),s(ye),s(we),s(S)),s(T),w(J,e),w(u,e),w(d,e),w(b,e),w(I,e),w(k,e),w($,e),w(B,e),w(R,e),w(Z,e),w(z,e),w(x,e)}}}const xe='{"title":"Papers, related resources & how to cite","local":"papers-related-resources--how-to-cite","sections":[{"title":"SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression (Jun 2023)","local":"spqr-a-sparse-quantized-representation-for-near-lossless-llm-weight-compression-jun-2023","sections":[],"depth":2},{"title":"QLoRA: Efficient Finetuning of Quantized LLMs (May 2023)","local":"qlora-efficient-finetuning-of-quantized-llms-may-2023","sections":[],"depth":2},{"title":"The case for 4-bit precision: k-bit Inference Scaling Laws (Dec 2022)","local":"the-case-for-4-bit-precision-k-bit-inference-scaling-laws-dec-2022","sections":[],"depth":2},{"title":"LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale (Nov 2022)","local":"llm-int8","sections":[],"depth":2},{"title":"8-bit Optimizers via Block-wise Quantization (Oct 2021)","local":"8-bit-optimizers-via-block-wise-quantization-oct-2021","sections":[],"depth":2}],"depth":1}';function Ne(Te){return Ee(()=>{new URLSearchParams(window.location.search).get("fw")}),[]}class Fe extends Ae{constructor(T){super(),Re(this,T,Ne,ze,$e,{})}}export{Fe as component};