You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

index.html 40 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842
  1. <!doctype html>
  2. <html lang="en" class="no-js">
  3. <head>
  4. <meta charset="utf-8">
  5. <meta name="viewport" content="width=device-width,initial-scale=1">
  6. <link rel="prev" href="../Architecture/">
  7. <link rel="next" href="../ContributingGuide/">
  8. <link rel="icon" href="../assets/images/favicon.png">
  9. <meta name="generator" content="mkdocs-1.4.3, mkdocs-material-9.1.16">
  10. <title>Tricks for FAQ - LLamaSharp Documentation</title>
  11. <link rel="stylesheet" href="../assets/stylesheets/main.26e3688c.min.css">
  12. <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
  13. <link rel="stylesheet" href="https://fonts.googleapis.com/css?family=Roboto:300,300i,400,400i,700,700i%7CRoboto+Mono:400,400i,700,700i&display=fallback">
  14. <style>:root{--md-text-font:"Roboto";--md-code-font:"Roboto Mono"}</style>
  15. <script>__md_scope=new URL("..",location),__md_hash=e=>[...e].reduce((e,_)=>(e<<5)-e+_.charCodeAt(0),0),__md_get=(e,_=localStorage,t=__md_scope)=>JSON.parse(_.getItem(t.pathname+"."+e)),__md_set=(e,_,t=localStorage,a=__md_scope)=>{try{t.setItem(a.pathname+"."+e,JSON.stringify(_))}catch(e){}}</script>
  16. </head>
  17. <body dir="ltr">
  18. <script>var palette=__md_get("__palette");if(palette&&"object"==typeof palette.color)for(var key of Object.keys(palette.color))document.body.setAttribute("data-md-color-"+key,palette.color[key])</script>
  19. <input class="md-toggle" data-md-toggle="drawer" type="checkbox" id="__drawer" autocomplete="off">
  20. <input class="md-toggle" data-md-toggle="search" type="checkbox" id="__search" autocomplete="off">
  21. <label class="md-overlay" for="__drawer"></label>
  22. <div data-md-component="skip">
  23. <a href="#tricks-for-faq" class="md-skip">
  24. Skip to content
  25. </a>
  26. </div>
  27. <div data-md-component="announce">
  28. </div>
  29. <div data-md-color-scheme="default" data-md-component="outdated" hidden>
  30. </div>
  31. <header class="md-header md-header--shadow" data-md-component="header">
  32. <nav class="md-header__inner md-grid" aria-label="Header">
  33. <a href=".." title="LLamaSharp Documentation" class="md-header__button md-logo" aria-label="LLamaSharp Documentation" data-md-component="logo">
  34. <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M12 8a3 3 0 0 0 3-3 3 3 0 0 0-3-3 3 3 0 0 0-3 3 3 3 0 0 0 3 3m0 3.54C9.64 9.35 6.5 8 3 8v11c3.5 0 6.64 1.35 9 3.54 2.36-2.19 5.5-3.54 9-3.54V8c-3.5 0-6.64 1.35-9 3.54Z"/></svg>
  35. </a>
  36. <label class="md-header__button md-icon" for="__drawer">
  37. <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M3 6h18v2H3V6m0 5h18v2H3v-2m0 5h18v2H3v-2Z"/></svg>
  38. </label>
  39. <div class="md-header__title" data-md-component="header-title">
  40. <div class="md-header__ellipsis">
  41. <div class="md-header__topic">
  42. <span class="md-ellipsis">
  43. LLamaSharp Documentation
  44. </span>
  45. </div>
  46. <div class="md-header__topic" data-md-component="header-topic">
  47. <span class="md-ellipsis">
  48. Tricks for FAQ
  49. </span>
  50. </div>
  51. </div>
  52. </div>
  53. <label class="md-header__button md-icon" for="__search">
  54. <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.516 6.516 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5Z"/></svg>
  55. </label>
  56. <div class="md-search" data-md-component="search" role="dialog">
  57. <label class="md-search__overlay" for="__search"></label>
  58. <div class="md-search__inner" role="search">
  59. <form class="md-search__form" name="search">
  60. <input type="text" class="md-search__input" name="query" aria-label="Search" placeholder="Search" autocapitalize="off" autocorrect="off" autocomplete="off" spellcheck="false" data-md-component="search-query" required>
  61. <label class="md-search__icon md-icon" for="__search">
  62. <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.516 6.516 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5Z"/></svg>
  63. <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M20 11v2H8l5.5 5.5-1.42 1.42L4.16 12l7.92-7.92L13.5 5.5 8 11h12Z"/></svg>
  64. </label>
  65. <nav class="md-search__options" aria-label="Search">
  66. <button type="reset" class="md-search__icon md-icon" title="Clear" aria-label="Clear" tabindex="-1">
  67. <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M19 6.41 17.59 5 12 10.59 6.41 5 5 6.41 10.59 12 5 17.59 6.41 19 12 13.41 17.59 19 19 17.59 13.41 12 19 6.41Z"/></svg>
  68. </button>
  69. </nav>
  70. </form>
  71. <div class="md-search__output">
  72. <div class="md-search__scrollwrap" data-md-scrollfix>
  73. <div class="md-search-result" data-md-component="search-result">
  74. <div class="md-search-result__meta">
  75. Initializing search
  76. </div>
  77. <ol class="md-search-result__list" role="presentation"></ol>
  78. </div>
  79. </div>
  80. </div>
  81. </div>
  82. </div>
  83. </nav>
  84. </header>
  85. <div class="md-container" data-md-component="container">
  86. <main class="md-main" data-md-component="main">
  87. <div class="md-main__inner md-grid">
  88. <div class="md-sidebar md-sidebar--primary" data-md-component="sidebar" data-md-type="navigation" >
  89. <div class="md-sidebar__scrollwrap">
  90. <div class="md-sidebar__inner">
  91. <nav class="md-nav md-nav--primary" aria-label="Navigation" data-md-level="0">
  92. <label class="md-nav__title" for="__drawer">
  93. <a href=".." title="LLamaSharp Documentation" class="md-nav__button md-logo" aria-label="LLamaSharp Documentation" data-md-component="logo">
  94. <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M12 8a3 3 0 0 0 3-3 3 3 0 0 0-3-3 3 3 0 0 0-3 3 3 3 0 0 0 3 3m0 3.54C9.64 9.35 6.5 8 3 8v11c3.5 0 6.64 1.35 9 3.54 2.36-2.19 5.5-3.54 9-3.54V8c-3.5 0-6.64 1.35-9 3.54Z"/></svg>
  95. </a>
  96. LLamaSharp Documentation
  97. </label>
  98. <ul class="md-nav__list" data-md-scrollfix>
  99. <li class="md-nav__item">
  100. <a href=".." class="md-nav__link">
  101. Overview
  102. </a>
  103. </li>
  104. <li class="md-nav__item">
  105. <a href="../GetStarted/" class="md-nav__link">
  106. Get Started
  107. </a>
  108. </li>
  109. <li class="md-nav__item">
  110. <a href="../Architecture/" class="md-nav__link">
  111. Architecture
  112. </a>
  113. </li>
  114. <li class="md-nav__item md-nav__item--active">
  115. <input class="md-nav__toggle md-toggle" type="checkbox" id="__toc">
  116. <label class="md-nav__link md-nav__link--active" for="__toc">
  117. Tricks for FAQ
  118. <span class="md-nav__icon md-icon"></span>
  119. </label>
  120. <a href="./" class="md-nav__link md-nav__link--active">
  121. Tricks for FAQ
  122. </a>
  123. <nav class="md-nav md-nav--secondary" aria-label="Table of contents">
  124. <label class="md-nav__title" for="__toc">
  125. <span class="md-nav__icon md-icon"></span>
  126. Table of contents
  127. </label>
  128. <ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
  129. <li class="md-nav__item">
  130. <a href="#carefully-set-the-anti-prompts" class="md-nav__link">
  131. Carefully set the anti-prompts
  132. </a>
  133. </li>
  134. <li class="md-nav__item">
  135. <a href="#pay-attention-to-the-length-of-prompt" class="md-nav__link">
  136. Pay attention to the length of prompt
  137. </a>
  138. </li>
  139. <li class="md-nav__item">
  140. <a href="#try-differenct-executors-with-a-prompt" class="md-nav__link">
  141. Try differenct executors with a prompt
  142. </a>
  143. </li>
  144. <li class="md-nav__item">
  145. <a href="#choose-models-weight-depending-on-you-task" class="md-nav__link">
  146. Choose models weight depending on you task
  147. </a>
  148. </li>
  149. <li class="md-nav__item">
  150. <a href="#set-the-layer-count-you-want-to-offload-to-gpu" class="md-nav__link">
  151. Set the layer count you want to offload to GPU
  152. </a>
  153. </li>
  154. </ul>
  155. </nav>
  156. </li>
  157. <li class="md-nav__item">
  158. <a href="../ContributingGuide/" class="md-nav__link">
  159. Contributing Guide
  160. </a>
  161. </li>
  162. <li class="md-nav__item md-nav__item--nested">
  163. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_6" >
  164. <label class="md-nav__link" for="__nav_6" id="__nav_6_label" tabindex="0">
  165. LLamaModel
  166. <span class="md-nav__icon md-icon"></span>
  167. </label>
  168. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_6_label" aria-expanded="false">
  169. <label class="md-nav__title" for="__nav_6">
  170. <span class="md-nav__icon md-icon"></span>
  171. LLamaModel
  172. </label>
  173. <ul class="md-nav__list" data-md-scrollfix>
  174. <li class="md-nav__item">
  175. <a href="../LLamaModel/parameters/" class="md-nav__link">
  176. Model Parameters
  177. </a>
  178. </li>
  179. <li class="md-nav__item">
  180. <a href="../LLamaModel/tokenization/" class="md-nav__link">
  181. Tokenization
  182. </a>
  183. </li>
  184. <li class="md-nav__item">
  185. <a href="../LLamaModel/embeddings/" class="md-nav__link">
  186. Get Embeddings
  187. </a>
  188. </li>
  189. <li class="md-nav__item">
  190. <a href="../LLamaModel/quantization/" class="md-nav__link">
  191. Quantization
  192. </a>
  193. </li>
  194. <li class="md-nav__item">
  195. <a href="../LLamaModel/save-load-state/" class="md-nav__link">
  196. Save/Load State
  197. </a>
  198. </li>
  199. </ul>
  200. </nav>
  201. </li>
  202. <li class="md-nav__item md-nav__item--nested">
  203. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_7" >
  204. <label class="md-nav__link" for="__nav_7" id="__nav_7_label" tabindex="0">
  205. LLamaExecutors
  206. <span class="md-nav__icon md-icon"></span>
  207. </label>
  208. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_7_label" aria-expanded="false">
  209. <label class="md-nav__title" for="__nav_7">
  210. <span class="md-nav__icon md-icon"></span>
  211. LLamaExecutors
  212. </label>
  213. <ul class="md-nav__list" data-md-scrollfix>
  214. <li class="md-nav__item">
  215. <a href="../LLamaExecutors/parameters/" class="md-nav__link">
  216. Inference Parameters
  217. </a>
  218. </li>
  219. <li class="md-nav__item">
  220. <a href="../LLamaExecutors/text-to-text-apis/" class="md-nav__link">
  221. Text-to-Text APIs
  222. </a>
  223. </li>
  224. <li class="md-nav__item">
  225. <a href="../LLamaExecutors/save-load-state/" class="md-nav__link">
  226. Save/Load State
  227. </a>
  228. </li>
  229. <li class="md-nav__item">
  230. <a href="../LLamaExecutors/differences/" class="md-nav__link">
  231. Differences of Executors
  232. </a>
  233. </li>
  234. </ul>
  235. </nav>
  236. </li>
  237. <li class="md-nav__item md-nav__item--nested">
  238. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_8" >
  239. <label class="md-nav__link" for="__nav_8" id="__nav_8_label" tabindex="0">
  240. ChatSession
  241. <span class="md-nav__icon md-icon"></span>
  242. </label>
  243. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_8_label" aria-expanded="false">
  244. <label class="md-nav__title" for="__nav_8">
  245. <span class="md-nav__icon md-icon"></span>
  246. ChatSession
  247. </label>
  248. <ul class="md-nav__list" data-md-scrollfix>
  249. <li class="md-nav__item">
  250. <a href="../ChatSession/basic-usages/" class="md-nav__link">
  251. Basic Usages
  252. </a>
  253. </li>
  254. <li class="md-nav__item">
  255. <a href="../ChatSession/transforms/" class="md-nav__link">
  256. Transoforms
  257. </a>
  258. </li>
  259. <li class="md-nav__item">
  260. <a href="../ChatSession/save-load-session/" class="md-nav__link">
  261. Save/Load Session
  262. </a>
  263. </li>
  264. </ul>
  265. </nav>
  266. </li>
  267. <li class="md-nav__item md-nav__item--nested">
  268. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_9" >
  269. <label class="md-nav__link" for="__nav_9" id="__nav_9_label" tabindex="0">
  270. Non-English Usages
  271. <span class="md-nav__icon md-icon"></span>
  272. </label>
  273. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_9_label" aria-expanded="false">
  274. <label class="md-nav__title" for="__nav_9">
  275. <span class="md-nav__icon md-icon"></span>
  276. Non-English Usages
  277. </label>
  278. <ul class="md-nav__list" data-md-scrollfix>
  279. <li class="md-nav__item">
  280. <a href="../NonEnglishUsage/Chinese/" class="md-nav__link">
  281. Chinese
  282. </a>
  283. </li>
  284. </ul>
  285. </nav>
  286. </li>
  287. <li class="md-nav__item md-nav__item--nested">
  288. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_10" >
  289. <label class="md-nav__link" for="__nav_10" id="__nav_10_label" tabindex="0">
  290. High-level Applications
  291. <span class="md-nav__icon md-icon"></span>
  292. </label>
  293. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_10_label" aria-expanded="false">
  294. <label class="md-nav__title" for="__nav_10">
  295. <span class="md-nav__icon md-icon"></span>
  296. High-level Applications
  297. </label>
  298. <ul class="md-nav__list" data-md-scrollfix>
  299. <li class="md-nav__item">
  300. <a href="../HighLevelApps/bot-sharp/" class="md-nav__link">
  301. BotSharp
  302. </a>
  303. </li>
  304. </ul>
  305. </nav>
  306. </li>
  307. <li class="md-nav__item md-nav__item--nested">
  308. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_11" >
  309. <label class="md-nav__link" for="__nav_11" id="__nav_11_label" tabindex="0">
  310. More
  311. <span class="md-nav__icon md-icon"></span>
  312. </label>
  313. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_11_label" aria-expanded="false">
  314. <label class="md-nav__title" for="__nav_11">
  315. <span class="md-nav__icon md-icon"></span>
  316. More
  317. </label>
  318. <ul class="md-nav__list" data-md-scrollfix>
  319. <li class="md-nav__item">
  320. <a href="../More/log/" class="md-nav__link">
  321. Logger
  322. </a>
  323. </li>
  324. </ul>
  325. </nav>
  326. </li>
  327. <li class="md-nav__item md-nav__item--nested">
  328. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_12" >
  329. <label class="md-nav__link" for="__nav_12" id="__nav_12_label" tabindex="0">
  330. Examples
  331. <span class="md-nav__icon md-icon"></span>
  332. </label>
  333. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_12_label" aria-expanded="false">
  334. <label class="md-nav__title" for="__nav_12">
  335. <span class="md-nav__icon md-icon"></span>
  336. Examples
  337. </label>
  338. <ul class="md-nav__list" data-md-scrollfix>
  339. <li class="md-nav__item">
  340. <a href="../Examples/ChatSessionStripRoleName/" class="md-nav__link">
  341. Chat session 1
  342. </a>
  343. </li>
  344. <li class="md-nav__item">
  345. <a href="../Examples/ChatSessionWithRoleName/" class="md-nav__link">
  346. Chat session 2
  347. </a>
  348. </li>
  349. <li class="md-nav__item">
  350. <a href="../Examples/GetEmbeddings/" class="md-nav__link">
  351. Get embeddings
  352. </a>
  353. </li>
  354. <li class="md-nav__item">
  355. <a href="../Examples/InstructModeExecute/" class="md-nav__link">
  356. Instruct executor
  357. </a>
  358. </li>
  359. <li class="md-nav__item">
  360. <a href="../Examples/InteractiveModeExecute/" class="md-nav__link">
  361. Interactive executor
  362. </a>
  363. </li>
  364. <li class="md-nav__item">
  365. <a href="../Examples/StatelessModeExecute/" class="md-nav__link">
  366. Stateless exeutor
  367. </a>
  368. </li>
  369. <li class="md-nav__item">
  370. <a href="../Examples/LoadAndSaveSession/" class="md-nav__link">
  371. Load/Save session
  372. </a>
  373. </li>
  374. <li class="md-nav__item">
  375. <a href="../Examples/LoadAndSaveState/" class="md-nav__link">
  376. Load/Save state
  377. </a>
  378. </li>
  379. <li class="md-nav__item">
  380. <a href="../Examples/QuantizeModel/" class="md-nav__link">
  381. Quantize model
  382. </a>
  383. </li>
  384. </ul>
  385. </nav>
  386. </li>
  387. <li class="md-nav__item md-nav__item--nested">
  388. <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_13" >
  389. <label class="md-nav__link" for="__nav_13" id="__nav_13_label" tabindex="0">
  390. API Reference
  391. <span class="md-nav__icon md-icon"></span>
  392. </label>
  393. <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_13_label" aria-expanded="false">
  394. <label class="md-nav__title" for="__nav_13">
  395. <span class="md-nav__icon md-icon"></span>
  396. API Reference
  397. </label>
  398. <ul class="md-nav__list" data-md-scrollfix>
  399. <li class="md-nav__item">
  400. <a href="../xmldocs/" class="md-nav__link">
  401. index
  402. </a>
  403. </li>
  404. <li class="md-nav__item">
  405. <a href="../xmldocs/llama.abstractions.ihistorytransform/" class="md-nav__link">
  406. llama.abstractions.ihistorytransform
  407. </a>
  408. </li>
  409. <li class="md-nav__item">
  410. <a href="../xmldocs/llama.abstractions.illamaexecutor/" class="md-nav__link">
  411. llama.abstractions.illamaexecutor
  412. </a>
  413. </li>
  414. <li class="md-nav__item">
  415. <a href="../xmldocs/llama.abstractions.itextstreamtransform/" class="md-nav__link">
  416. llama.abstractions.itextstreamtransform
  417. </a>
  418. </li>
  419. <li class="md-nav__item">
  420. <a href="../xmldocs/llama.abstractions.itexttransform/" class="md-nav__link">
  421. llama.abstractions.itexttransform
  422. </a>
  423. </li>
  424. <li class="md-nav__item">
  425. <a href="../xmldocs/llama.chatsession/" class="md-nav__link">
  426. llama.chatsession
  427. </a>
  428. </li>
  429. <li class="md-nav__item">
  430. <a href="../xmldocs/llama.common.authorrole/" class="md-nav__link">
  431. llama.common.authorrole
  432. </a>
  433. </li>
  434. <li class="md-nav__item">
  435. <a href="../xmldocs/llama.common.chathistory/" class="md-nav__link">
  436. llama.common.chathistory
  437. </a>
  438. </li>
  439. <li class="md-nav__item">
  440. <a href="../xmldocs/llama.common.fixedsizequeue-1/" class="md-nav__link">
  441. llama.common.fixedsizequeue-1
  442. </a>
  443. </li>
  444. <li class="md-nav__item">
  445. <a href="../xmldocs/llama.common.illamalogger/" class="md-nav__link">
  446. llama.common.illamalogger
  447. </a>
  448. </li>
  449. <li class="md-nav__item">
  450. <a href="../xmldocs/llama.common.inferenceparams/" class="md-nav__link">
  451. llama.common.inferenceparams
  452. </a>
  453. </li>
  454. <li class="md-nav__item">
  455. <a href="../xmldocs/llama.common.llamadefaultlogger/" class="md-nav__link">
  456. llama.common.llamadefaultlogger
  457. </a>
  458. </li>
  459. <li class="md-nav__item">
  460. <a href="../xmldocs/llama.common.mirostatetype/" class="md-nav__link">
  461. llama.common.mirostatetype
  462. </a>
  463. </li>
  464. <li class="md-nav__item">
  465. <a href="../xmldocs/llama.common.modelparams/" class="md-nav__link">
  466. llama.common.modelparams
  467. </a>
  468. </li>
  469. <li class="md-nav__item">
  470. <a href="../xmldocs/llama.exceptions.runtimeerror/" class="md-nav__link">
  471. llama.exceptions.runtimeerror
  472. </a>
  473. </li>
  474. <li class="md-nav__item">
  475. <a href="../xmldocs/llama.extensions.dictionaryextension/" class="md-nav__link">
  476. llama.extensions.dictionaryextension
  477. </a>
  478. </li>
  479. <li class="md-nav__item">
  480. <a href="../xmldocs/llama.instructexecutor/" class="md-nav__link">
  481. llama.instructexecutor
  482. </a>
  483. </li>
  484. <li class="md-nav__item">
  485. <a href="../xmldocs/llama.interactiveexecutor/" class="md-nav__link">
  486. llama.interactiveexecutor
  487. </a>
  488. </li>
  489. <li class="md-nav__item">
  490. <a href="../xmldocs/llama.llamaembedder/" class="md-nav__link">
  491. llama.llamaembedder
  492. </a>
  493. </li>
  494. <li class="md-nav__item">
  495. <a href="../xmldocs/llama.llamamodel/" class="md-nav__link">
  496. llama.llamamodel
  497. </a>
  498. </li>
  499. <li class="md-nav__item">
  500. <a href="../xmldocs/llama.llamaquantizer/" class="md-nav__link">
  501. llama.llamaquantizer
  502. </a>
  503. </li>
  504. <li class="md-nav__item">
  505. <a href="../xmldocs/llama.llamatransforms/" class="md-nav__link">
  506. llama.llamatransforms
  507. </a>
  508. </li>
  509. <li class="md-nav__item">
  510. <a href="../xmldocs/llama.native.llamacontextparams/" class="md-nav__link">
  511. llama.native.llamacontextparams
  512. </a>
  513. </li>
  514. <li class="md-nav__item">
  515. <a href="../xmldocs/llama.native.llamaftype/" class="md-nav__link">
  516. llama.native.llamaftype
  517. </a>
  518. </li>
  519. <li class="md-nav__item">
  520. <a href="../xmldocs/llama.native.llamatokendata/" class="md-nav__link">
  521. llama.native.llamatokendata
  522. </a>
  523. </li>
  524. <li class="md-nav__item">
  525. <a href="../xmldocs/llama.native.llamatokendataarray/" class="md-nav__link">
  526. llama.native.llamatokendataarray
  527. </a>
  528. </li>
  529. <li class="md-nav__item">
  530. <a href="../xmldocs/llama.native.llamatokendataarraynative/" class="md-nav__link">
  531. llama.native.llamatokendataarraynative
  532. </a>
  533. </li>
  534. <li class="md-nav__item">
  535. <a href="../xmldocs/llama.native.nativeapi/" class="md-nav__link">
  536. llama.native.nativeapi
  537. </a>
  538. </li>
  539. <li class="md-nav__item">
  540. <a href="../xmldocs/llama.native.safellamacontexthandle/" class="md-nav__link">
  541. llama.native.safellamacontexthandle
  542. </a>
  543. </li>
  544. <li class="md-nav__item">
  545. <a href="../xmldocs/llama.native.safellamahandlebase/" class="md-nav__link">
  546. llama.native.safellamahandlebase
  547. </a>
  548. </li>
  549. <li class="md-nav__item">
  550. <a href="../xmldocs/llama.oldversion.chatcompletion/" class="md-nav__link">
  551. llama.oldversion.chatcompletion
  552. </a>
  553. </li>
  554. <li class="md-nav__item">
  555. <a href="../xmldocs/llama.oldversion.chatcompletionchoice/" class="md-nav__link">
  556. llama.oldversion.chatcompletionchoice
  557. </a>
  558. </li>
  559. <li class="md-nav__item">
  560. <a href="../xmldocs/llama.oldversion.chatcompletionchunk/" class="md-nav__link">
  561. llama.oldversion.chatcompletionchunk
  562. </a>
  563. </li>
  564. <li class="md-nav__item">
  565. <a href="../xmldocs/llama.oldversion.chatcompletionchunkchoice/" class="md-nav__link">
  566. llama.oldversion.chatcompletionchunkchoice
  567. </a>
  568. </li>
  569. <li class="md-nav__item">
  570. <a href="../xmldocs/llama.oldversion.chatcompletionchunkdelta/" class="md-nav__link">
  571. llama.oldversion.chatcompletionchunkdelta
  572. </a>
  573. </li>
  574. <li class="md-nav__item">
  575. <a href="../xmldocs/llama.oldversion.chatcompletionmessage/" class="md-nav__link">
  576. llama.oldversion.chatcompletionmessage
  577. </a>
  578. </li>
  579. <li class="md-nav__item">
  580. <a href="../xmldocs/llama.oldversion.chatmessagerecord/" class="md-nav__link">
  581. llama.oldversion.chatmessagerecord
  582. </a>
  583. </li>
  584. <li class="md-nav__item">
  585. <a href="../xmldocs/llama.oldversion.chatrole/" class="md-nav__link">
  586. llama.oldversion.chatrole
  587. </a>
  588. </li>
  589. <li class="md-nav__item">
  590. <a href="../xmldocs/llama.oldversion.chatsession-1/" class="md-nav__link">
  591. llama.oldversion.chatsession-1
  592. </a>
  593. </li>
  594. <li class="md-nav__item">
  595. <a href="../xmldocs/llama.oldversion.completion/" class="md-nav__link">
  596. llama.oldversion.completion
  597. </a>
  598. </li>
  599. <li class="md-nav__item">
  600. <a href="../xmldocs/llama.oldversion.completionchoice/" class="md-nav__link">
  601. llama.oldversion.completionchoice
  602. </a>
  603. </li>
  604. <li class="md-nav__item">
  605. <a href="../xmldocs/llama.oldversion.completionchunk/" class="md-nav__link">
  606. llama.oldversion.completionchunk
  607. </a>
  608. </li>
  609. <li class="md-nav__item">
  610. <a href="../xmldocs/llama.oldversion.completionlogprobs/" class="md-nav__link">
  611. llama.oldversion.completionlogprobs
  612. </a>
  613. </li>
  614. <li class="md-nav__item">
  615. <a href="../xmldocs/llama.oldversion.completionusage/" class="md-nav__link">
  616. llama.oldversion.completionusage
  617. </a>
  618. </li>
  619. <li class="md-nav__item">
  620. <a href="../xmldocs/llama.oldversion.embedding/" class="md-nav__link">
  621. llama.oldversion.embedding
  622. </a>
  623. </li>
  624. <li class="md-nav__item">
  625. <a href="../xmldocs/llama.oldversion.embeddingdata/" class="md-nav__link">
  626. llama.oldversion.embeddingdata
  627. </a>
  628. </li>
  629. <li class="md-nav__item">
  630. <a href="../xmldocs/llama.oldversion.embeddingusage/" class="md-nav__link">
  631. llama.oldversion.embeddingusage
  632. </a>
  633. </li>
  634. <li class="md-nav__item">
  635. <a href="../xmldocs/llama.oldversion.ichatmodel/" class="md-nav__link">
  636. llama.oldversion.ichatmodel
  637. </a>
  638. </li>
  639. <li class="md-nav__item">
  640. <a href="../xmldocs/llama.oldversion.llamaembedder/" class="md-nav__link">
  641. llama.oldversion.llamaembedder
  642. </a>
  643. </li>
  644. <li class="md-nav__item">
  645. <a href="../xmldocs/llama.oldversion.llamamodel/" class="md-nav__link">
  646. llama.oldversion.llamamodel
  647. </a>
  648. </li>
  649. <li class="md-nav__item">
  650. <a href="../xmldocs/llama.oldversion.llamaparams/" class="md-nav__link">
  651. llama.oldversion.llamaparams
  652. </a>
  653. </li>
  654. <li class="md-nav__item">
  655. <a href="../xmldocs/llama.resettablellamamodel/" class="md-nav__link">
  656. llama.resettablellamamodel
  657. </a>
  658. </li>
  659. <li class="md-nav__item">
  660. <a href="../xmldocs/llama.statefulexecutorbase/" class="md-nav__link">
  661. llama.statefulexecutorbase
  662. </a>
  663. </li>
  664. <li class="md-nav__item">
  665. <a href="../xmldocs/llama.statelessexecutor/" class="md-nav__link">
  666. llama.statelessexecutor
  667. </a>
  668. </li>
  669. </ul>
  670. </nav>
  671. </li>
  672. </ul>
  673. </nav>
  674. </div>
  675. </div>
  676. </div>
  677. <div class="md-sidebar md-sidebar--secondary" data-md-component="sidebar" data-md-type="toc" >
  678. <div class="md-sidebar__scrollwrap">
  679. <div class="md-sidebar__inner">
  680. <nav class="md-nav md-nav--secondary" aria-label="Table of contents">
  681. <label class="md-nav__title" for="__toc">
  682. <span class="md-nav__icon md-icon"></span>
  683. Table of contents
  684. </label>
  685. <ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
  686. <li class="md-nav__item">
  687. <a href="#carefully-set-the-anti-prompts" class="md-nav__link">
  688. Carefully set the anti-prompts
  689. </a>
  690. </li>
  691. <li class="md-nav__item">
  692. <a href="#pay-attention-to-the-length-of-prompt" class="md-nav__link">
  693. Pay attention to the length of prompt
  694. </a>
  695. </li>
  696. <li class="md-nav__item">
  697. <a href="#try-differenct-executors-with-a-prompt" class="md-nav__link">
  698. Try differenct executors with a prompt
  699. </a>
  700. </li>
  701. <li class="md-nav__item">
  702. <a href="#choose-models-weight-depending-on-you-task" class="md-nav__link">
  703. Choose models weight depending on you task
  704. </a>
  705. </li>
  706. <li class="md-nav__item">
  707. <a href="#set-the-layer-count-you-want-to-offload-to-gpu" class="md-nav__link">
  708. Set the layer count you want to offload to GPU
  709. </a>
  710. </li>
  711. </ul>
  712. </nav>
  713. </div>
  714. </div>
  715. </div>
  716. <div class="md-content" data-md-component="content">
  717. <article class="md-content__inner md-typeset">
  718. <h1 id="tricks-for-faq">Tricks for FAQ</h1>
  719. <p>Sometimes, your application with LLM and LLamaSharp may have strange behaviors. Before opening an issue to report the BUG, the following tricks may worth a try.</p>
  720. <h2 id="carefully-set-the-anti-prompts">Carefully set the anti-prompts</h2>
  721. <p>Anti-prompt can also be called as "Stop-keyword", which decides when to stop the response generation. Under interactive mode, the maximum tokens count is always not set, which makes the LLM generates responses infinitively. Therefore, setting anti-prompt correctly helps a lot to avoid the strange behaviors. For example, the prompt file <code>chat-with-bob.txt</code> has the following content:</p>
  722. <pre><code>Transcript of a dialog, where the User interacts with an Assistant named Bob. Bob is helpful, kind, honest, good at writing, and never fails to answer the User's requests immediately and with precision.
  723. User: Hello, Bob.
  724. Bob: Hello. How may I help you today?
  725. User: Please tell me the largest city in Europe.
  726. Bob: Sure. The largest city in Europe is Moscow, the capital of Russia.
  727. User:
  728. </code></pre>
  729. <p>Therefore, the anti-prompt should be set as "User:". If the last line of the prompt is removed, LLM will automatically generate a question (user) and a response (bob) for one time when running the chat session. Therefore, the antiprompt is suggested to be appended to the prompt when starting a chat session.</p>
  730. <p>What if an extra line is appended? The string "User:" in the prompt will be followed with a char "\n". Thus when running the model, the automatic generation of a pair of question and response may appear because the anti-prompt is "User:" but the last token is "User:\n". As for whether it will appear, it's an undefined behavior, which depends on the implementation inside the <code>LLamaExecutor</code>. Anyway, since it may leads to unexpected behaviors, it's recommended to trim your prompt or carefully keep consistent with your anti-prompt.</p>
  731. <h2 id="pay-attention-to-the-length-of-prompt">Pay attention to the length of prompt</h2>
  732. <p>Sometimes we want to input a long prompt to execute a task. However, the context size may limit the inference of LLama model. Please ensure the inequality below holds.</p>
  733. <p>$$ len(prompt) + len(response) &lt; len(context) $$</p>
  734. <p>In this inequality, <code>len(response)</code> refers to the expected tokens for LLM to generate.</p>
  735. <h2 id="try-differenct-executors-with-a-prompt">Try differenct executors with a prompt</h2>
  736. <p>Some prompt works well under interactive mode, such as <code>chat-with-bob</code>, some others may work well with instruct mode, such as <code>alpaca</code>. Besides, if your input is quite simple and one-time job, such as "Q: what is the satellite of the earth? A: ", stateless mode will be a good choice.</p>
  737. <p>If your chat bot has bad performance, trying different executor will possibly make it work well.</p>
  738. <h2 id="choose-models-weight-depending-on-you-task">Choose models weight depending on you task</h2>
  739. <p>The differences between modes may lead to much different behaviors under the same task. For example, if you're building a chat bot with non-English, a fine-tuned model specially for the language you want to use will have huge effect on the performance.</p>
  740. <h2 id="set-the-layer-count-you-want-to-offload-to-gpu">Set the layer count you want to offload to GPU</h2>
  741. <p>Currently, the <code>GpuLayerCount</code> param, which decides the number of layer loaded into GPU, is set to 20 by default. However, if you have some efficient GPUs, setting it as a larger number will attain faster inference.</p>
  742. </article>
  743. </div>
  744. </div>
  745. </main>
  746. <footer class="md-footer">
  747. <div class="md-footer-meta md-typeset">
  748. <div class="md-footer-meta__inner md-grid">
  749. <div class="md-copyright">
  750. Made with
  751. <a href="https://squidfunk.github.io/mkdocs-material/" target="_blank" rel="noopener">
  752. Material for MkDocs
  753. </a>
  754. </div>
  755. </div>
  756. </div>
  757. </footer>
  758. </div>
  759. <div class="md-dialog" data-md-component="dialog">
  760. <div class="md-dialog__inner md-typeset"></div>
  761. </div>
  762. <script id="__config" type="application/json">{"base": "..", "features": [], "search": "../assets/javascripts/workers/search.208ed371.min.js", "translations": {"clipboard.copied": "Copied to clipboard", "clipboard.copy": "Copy to clipboard", "search.result.more.one": "1 more on this page", "search.result.more.other": "# more on this page", "search.result.none": "No matching documents", "search.result.one": "1 matching document", "search.result.other": "# matching documents", "search.result.placeholder": "Type to start searching", "search.result.term.missing": "Missing", "select.version": "Select version"}, "version": {"provider": "mike"}}</script>
  763. <script src="../assets/javascripts/bundle.a51614de.min.js"></script>
  764. </body>
  765. </html>

C#/.NET上易用的LLM高性能推理框架,支持LLaMA和LLaVA系列模型。

Contributors (1)