bitsandbytes/pr_2036/en/_app/immutable/nodes/8.D4eEyIB2.js

import"../chunks/DsnmJJEf.js";import{i as p,h as u,C as h,H as r,E as f,s as g}from"../chunks/CHailA3B.js";import{p as y,o as v,s as a,f as w,a as l,b as q,c as d,n as x}from"../chunks/CMkz9frW.js";const z='{"title":"bitsandbytes","local":"bitsandbytes","sections":[],"depth":1}';var L=d('<meta name="hf:doc:metadata"/>'),_=d("<p></p> <!> <!> <p>bitsandbytes enables accessible large language models via k-bit quantization for PyTorch. bitsandbytes provides three main features for dramatically reducing memory consumption for inference and training:</p> <ul><li>8-bit optimizers uses block-wise quantization to maintain 32-bit performance at a small fraction of the memory cost.</li> <li>LLM.int8() or 8-bit quantization enables large language model inference with only half the required memory and without any performance degradation. This method is based on vector-wise quantization to quantize most features to 8-bits and separately treating outliers with 16-bit matrix multiplication.</li> <li>QLoRA or 4-bit quantization enables large language model training with several memory-saving techniques that don’t compromise performance. This method quantizes a model to 4-bits and inserts a small set of trainable low-rank adaptation (LoRA) weights to allow training.</li></ul> <!> <p>bitsandbytes is MIT licensed.</p> <!> <p></p>",1);function k(m,c){y(c,!1),v(()=>{new URLSearchParams(window.location.search).get("fw")}),p();var t=_();u("1ldsu9f",o=>{var s=L();g(s,"content",z),l(o,s)});var e=a(w(t),2);h(e,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var i=a(e,2);r(i,{title:"bitsandbytes",local:"bitsandbytes",headingTag:"h1"});var n=a(i,6);r(n,{title:"License",local:"license",headingTag:"h1"});var b=a(n,4);f(b,{source:"https://github.com/bitsandbytes-foundation/bitsandbytes/blob/main/docs/source/index.mdx"}),x(2),l(m,t),q()}export{k as component};