diff --git a/.nojekyll b/.nojekyll
index 861de3733..003beb4a9 100644
--- a/.nojekyll
+++ b/.nojekyll
@@ -1 +1 @@
-f561c4bc
\ No newline at end of file
+b206da03
\ No newline at end of file
diff --git a/FAQS.html b/FAQS.html
index 6caab1255..5f8a998ee 100644
--- a/FAQS.html
+++ b/FAQS.html
@@ -183,8 +183,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="./docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="./docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -195,8 +195,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="./docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="./docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/TODO.html b/TODO.html
index 15907fa33..df26757a3 100644
--- a/TODO.html
+++ b/TODO.html
@@ -183,8 +183,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="./docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="./docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -195,8 +195,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="./docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="./docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/batch_vs_grad.html b/docs/batch_vs_grad.html
new file mode 100644
index 000000000..69ceaa53a
--- /dev/null
+++ b/docs/batch_vs_grad.html
@@ -0,0 +1,730 @@
+<!DOCTYPE html>
+<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en"><head>
+
+<meta charset="utf-8">
+<meta name="generator" content="quarto-1.4.553">
+
+<meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
+
+<meta name="description" content="Understanding of batch size and gradient accumulation steps">
+
+<title>Axolotl - Batch size vs Gradient accumulation</title>
+<style>
+code{white-space: pre-wrap;}
+span.smallcaps{font-variant: small-caps;}
+div.columns{display: flex; gap: min(4vw, 1.5em);}
+div.column{flex: auto; overflow-x: auto;}
+div.hanging-indent{margin-left: 1.5em; text-indent: -1.5em;}
+ul.task-list{list-style: none;}
+ul.task-list li input[type="checkbox"] {
+  width: 0.8em;
+  margin: 0 0.8em 0.2em -1em; /* quarto-specific, see https://github.com/quarto-dev/quarto-cli/issues/4556 */ 
+  vertical-align: middle;
+}
+</style>
+
+
+<script src="../site_libs/quarto-nav/quarto-nav.js"></script>
+<script src="../site_libs/clipboard/clipboard.min.js"></script>
+<script src="../site_libs/quarto-search/autocomplete.umd.js"></script>
+<script src="../site_libs/quarto-search/fuse.min.js"></script>
+<script src="../site_libs/quarto-search/quarto-search.js"></script>
+<meta name="quarto:offset" content="../">
+<link href="../favicon.jpg" rel="icon" type="image/jpeg">
+<script src="../site_libs/quarto-html/quarto.js"></script>
+<script src="../site_libs/quarto-html/popper.min.js"></script>
+<script src="../site_libs/quarto-html/tippy.umd.min.js"></script>
+<script src="../site_libs/quarto-html/anchor.min.js"></script>
+<link href="../site_libs/quarto-html/tippy.css" rel="stylesheet">
+<link href="../site_libs/quarto-html/quarto-syntax-highlighting.css" rel="stylesheet" id="quarto-text-highlighting-styles">
+<script src="../site_libs/bootstrap/bootstrap.min.js"></script>
+<link href="../site_libs/bootstrap/bootstrap-icons.css" rel="stylesheet">
+<link href="../site_libs/bootstrap/bootstrap.min.css" rel="stylesheet" id="quarto-bootstrap" data-mode="light">
+<script id="quarto-search-options" type="application/json">{
+  "location": "navbar",
+  "copy-button": false,
+  "collapse-after": 3,
+  "panel-placement": "end",
+  "type": "overlay",
+  "limit": 50,
+  "keyboard-shortcut": [
+    "f",
+    "/",
+    "s"
+  ],
+  "show-item-context": false,
+  "language": {
+    "search-no-results-text": "No results",
+    "search-matching-documents-text": "matching documents",
+    "search-copy-link-title": "Copy link to search",
+    "search-hide-matches-text": "Hide additional matches",
+    "search-more-match-text": "more match in this document",
+    "search-more-matches-text": "more matches in this document",
+    "search-clear-button-title": "Clear",
+    "search-text-placeholder": "",
+    "search-detached-cancel-button-title": "Cancel",
+    "search-submit-button-title": "Submit",
+    "search-label": "Search"
+  }
+}</script>
+
+
+<link rel="stylesheet" href="../styles.css">
+</head>
+
+<body class="nav-sidebar docked nav-fixed">
+
+<div id="quarto-search-results"></div>
+  <header id="quarto-header" class="headroom fixed-top">
+    <nav class="navbar navbar-expand " data-bs-theme="dark">
+      <div class="navbar-container container-fluid">
+      <div class="navbar-brand-container mx-auto">
+    <a class="navbar-brand" href="../index.html">
+    <span class="navbar-title">Axolotl</span>
+    </a>
+  </div>
+        <div class="quarto-navbar-tools tools-wide tools-end">
+    <a href="https://twitter.com/axolotl_ai" title="" class="quarto-navigation-tool px-1" aria-label=""><i class="bi bi-twitter"></i></a>
+    <a href="https://github.com/OpenAccess-AI-Collective/axolotl/" title="" class="quarto-navigation-tool px-1" aria-label=""><i class="bi bi-github"></i></a>
+    <a href="https://discord.gg/7m9sfhzaf3" title="" class="quarto-navigation-tool px-1" aria-label=""><i class="bi bi-discord"></i></a>
+</div>
+          <div id="quarto-search" class="" title="Search"></div>
+      </div> <!-- /container-fluid -->
+    </nav>
+  <nav class="quarto-secondary-nav">
+    <div class="container-fluid d-flex">
+      <button type="button" class="quarto-btn-toggle btn" data-bs-toggle="collapse" data-bs-target=".quarto-sidebar-collapse-item" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">
+        <i class="bi bi-layout-text-sidebar-reverse"></i>
+      </button>
+        <nav class="quarto-page-breadcrumbs" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item">Batch size vs Gradient accumulation</li></ol></nav>
+        <a class="flex-grow-1" role="button" data-bs-toggle="collapse" data-bs-target=".quarto-sidebar-collapse-item" aria-controls="quarto-sidebar" aria-expanded="false" aria-label="Toggle sidebar navigation" onclick="if (window.quartoToggleHeadroom) { window.quartoToggleHeadroom(); }">      
+        </a>
+    </div>
+  </nav>
+</header>
+<!-- content -->
+<div id="quarto-content" class="quarto-container page-columns page-rows-contents page-layout-article page-navbar">
+<!-- sidebar -->
+  <nav id="quarto-sidebar" class="sidebar collapse collapse-horizontal quarto-sidebar-collapse-item sidebar-navigation docked overflow-auto">
+    <div class="sidebar-menu-container"> 
+    <ul class="list-unstyled mt-1">
+        <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../index.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Home</span></a>
+  </div>
+</li>
+        <li class="sidebar-item sidebar-item-section">
+      <div class="sidebar-item-container"> 
+            <a class="sidebar-item-text sidebar-link text-start" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar-section-1" aria-expanded="true">
+ <span class="menu-text">How-To Guides</span></a>
+          <a class="sidebar-item-toggle text-start" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar-section-1" aria-expanded="true" aria-label="Toggle section">
+            <i class="bi bi-chevron-right ms-2"></i>
+          </a> 
+      </div>
+      <ul id="quarto-sidebar-section-1" class="collapse list-unstyled sidebar-section depth1 show">  
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/debugging.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Debugging</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/multipack.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Multipack (Sample Packing)</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/fsdp_qlora.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">FDSP + QLoRA</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/input_output.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Template-free prompt construction</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/rlhf.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">RLHF (Beta)</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/nccl.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">NCCL</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/mac.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Mac M-series</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/multi-node.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Multi Node</span></a>
+  </div>
+</li>
+      </ul>
+  </li>
+        <li class="sidebar-item sidebar-item-section">
+      <div class="sidebar-item-container"> 
+            <a href="../docs/dataset-formats/index.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Dataset Formats</span></a>
+          <a class="sidebar-item-toggle text-start" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar-section-2" aria-expanded="true" aria-label="Toggle section">
+            <i class="bi bi-chevron-right ms-2"></i>
+          </a> 
+      </div>
+      <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/dataset-formats/inst_tune.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Instruction Tuning</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/dataset-formats/template_free.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Template-Free</span></a>
+  </div>
+</li>
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/dataset-formats/tokenized.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Custom Pre-Tokenized Dataset</span></a>
+  </div>
+</li>
+      </ul>
+  </li>
+        <li class="sidebar-item sidebar-item-section">
+      <div class="sidebar-item-container"> 
+            <a class="sidebar-item-text sidebar-link text-start" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar-section-3" aria-expanded="true">
+ <span class="menu-text">Reference</span></a>
+          <a class="sidebar-item-toggle text-start" data-bs-toggle="collapse" data-bs-target="#quarto-sidebar-section-3" aria-expanded="true" aria-label="Toggle section">
+            <i class="bi bi-chevron-right ms-2"></i>
+          </a> 
+      </div>
+      <ul id="quarto-sidebar-section-3" class="collapse list-unstyled sidebar-section depth1 show">  
+          <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/config.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Config options</span></a>
+  </div>
+</li>
+      </ul>
+  </li>
+        <li class="sidebar-item">
+  <div class="sidebar-item-container"> 
+  <a href="../docs/faq.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">FAQ</span></a>
+  </div>
+</li>
+    </ul>
+    </div>
+</nav>
+<div id="quarto-sidebar-glass" class="quarto-sidebar-collapse-item" data-bs-toggle="collapse" data-bs-target=".quarto-sidebar-collapse-item"></div>
+<!-- margin-sidebar -->
+    <div id="quarto-margin-sidebar" class="sidebar margin-sidebar">
+        
+    </div>
+<!-- main -->
+<main class="content" id="quarto-document-content">
+
+<header id="title-block-header" class="quarto-title-block default">
+<div class="quarto-title">
+<h1 class="title">Batch size vs Gradient accumulation</h1>
+</div>
+
+<div>
+  <div class="description">
+    Understanding of batch size and gradient accumulation steps
+  </div>
+</div>
+
+
+<div class="quarto-title-meta">
+
+    
+  
+    
+  </div>
+  
+
+
+</header>
+
+
+<p>Gradient accumulation means accumulating gradients over several mini-batches and updating the model weights afterward. When the samples in each batch are diverse, this technique doesn’t significantly impact learning.</p>
+<p>This method allows for effective training with larger effective batch sizes without needing proportionally larger memory. Here’s why:</p>
+<ol type="1">
+<li><p><strong>Memory Consumption with Batch Size</strong>: The primary reason increasing the batch size impacts memory is due to the storage requirements for intermediate activations. When you forward propagate a batch through a network, you have to store the activations at each layer for each sample in the batch, because these activations are used during backpropagation to compute gradients. Therefore, larger batches mean more activations, leading to greater GPU memory consumption.</p></li>
+<li><p><strong>Gradient Accumulation</strong>: With gradient accumulation, you’re effectively simulating a larger batch size by accumulating gradients over several smaller batches (or micro-batches). However, at any given time, you’re only forward and backward propagating a micro-batch. This means you only store activations for the micro-batch, not the full accumulated batch. As a result, you can simulate the effect of a larger batch size without the memory cost of storing activations for a large batch.</p></li>
+</ol>
+<p><strong>Example 1:</strong> Micro batch size: 3 Gradient accumulation steps: 2 Number of GPUs: 3 Total batch size = 3 * 2 * 3 = 18</p>
+<pre><code>| GPU 1          | GPU 2          | GPU 3          |
+|----------------|----------------|----------------|
+| S1, S2, S3     | S4, S5, S6     | S7, S8, S9     |
+| e1, e2, e3     | e4, e5, e6     | e7, e8, e9     |
+|----------------|----------------|----------------|
+| → (accumulate) | → (accumulate) | → (accumulate) |
+|----------------|----------------|----------------|
+| S10, S11, S12  | S13, S14, S15  | S16, S17, S18  |
+| e10, e11, e12  | e13, e14, e15  | e16, e17, e18  |
+|----------------|----------------|----------------|
+| → (apply)      | → (apply)      | → (apply)      |
+
+Accumulated gradient for the weight w1 after the second iteration (considering all GPUs):
+Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6 + e7 + e8 + e9 + e10 + e11 + e12 + e13 + e14 + e15 + e16 + e17 + e18
+
+Weight update for w1:
+w1_new = w1_old - learning rate x (Total gradient for w1 / 18)</code></pre>
+<p><strong>Example 2:</strong> Micro batch size: 2 Gradient accumulation steps: 1 Number of GPUs: 3 Total batch size = 2 * 1 * 3 = 6</p>
+<pre><code>| GPU 1     | GPU 2     | GPU 3     |
+|-----------|-----------|-----------|
+| S1, S2    | S3, S4    | S5, S6    |
+| e1, e2    | e3, e4    | e5, e6    |
+|-----------|-----------|-----------|
+| → (apply) | → (apply) | → (apply) |
+
+Accumulated gradient for the weight w1 (considering all GPUs):
+Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6
+
+Weight update for w1:
+w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
+
+
+
+</main> <!-- /main -->
+<script id="quarto-html-after-body" type="application/javascript">
+window.document.addEventListener("DOMContentLoaded", function (event) {
+  const toggleBodyColorMode = (bsSheetEl) => {
+    const mode = bsSheetEl.getAttribute("data-mode");
+    const bodyEl = window.document.querySelector("body");
+    if (mode === "dark") {
+      bodyEl.classList.add("quarto-dark");
+      bodyEl.classList.remove("quarto-light");
+    } else {
+      bodyEl.classList.add("quarto-light");
+      bodyEl.classList.remove("quarto-dark");
+    }
+  }
+  const toggleBodyColorPrimary = () => {
+    const bsSheetEl = window.document.querySelector("link#quarto-bootstrap");
+    if (bsSheetEl) {
+      toggleBodyColorMode(bsSheetEl);
+    }
+  }
+  toggleBodyColorPrimary();  
+  const icon = "";
+  const anchorJS = new window.AnchorJS();
+  anchorJS.options = {
+    placement: 'right',
+    icon: icon
+  };
+  anchorJS.add('.anchored');
+  const isCodeAnnotation = (el) => {
+    for (const clz of el.classList) {
+      if (clz.startsWith('code-annotation-')) {                     
+        return true;
+      }
+    }
+    return false;
+  }
+  const clipboard = new window.ClipboardJS('.code-copy-button', {
+    text: function(trigger) {
+      const codeEl = trigger.previousElementSibling.cloneNode(true);
+      for (const childEl of codeEl.children) {
+        if (isCodeAnnotation(childEl)) {
+          childEl.remove();
+        }
+      }
+      return codeEl.innerText;
+    }
+  });
+  clipboard.on('success', function(e) {
+    // button target
+    const button = e.trigger;
+    // don't keep focus
+    button.blur();
+    // flash "checked"
+    button.classList.add('code-copy-button-checked');
+    var currentTitle = button.getAttribute("title");
+    button.setAttribute("title", "Copied!");
+    let tooltip;
+    if (window.bootstrap) {
+      button.setAttribute("data-bs-toggle", "tooltip");
+      button.setAttribute("data-bs-placement", "left");
+      button.setAttribute("data-bs-title", "Copied!");
+      tooltip = new bootstrap.Tooltip(button, 
+        { trigger: "manual", 
+          customClass: "code-copy-button-tooltip",
+          offset: [0, -8]});
+      tooltip.show();    
+    }
+    setTimeout(function() {
+      if (tooltip) {
+        tooltip.hide();
+        button.removeAttribute("data-bs-title");
+        button.removeAttribute("data-bs-toggle");
+        button.removeAttribute("data-bs-placement");
+      }
+      button.setAttribute("title", currentTitle);
+      button.classList.remove('code-copy-button-checked');
+    }, 1000);
+    // clear code selection
+    e.clearSelection();
+  });
+    var localhostRegex = new RegExp(/^(?:http|https):\/\/localhost\:?[0-9]*\//);
+    var mailtoRegex = new RegExp(/^mailto:/);
+      var filterRegex = new RegExp("https:\/\/OpenAccess-AI-Collective\.github\.io\/axolotl\/");
+    var isInternal = (href) => {
+        return filterRegex.test(href) || localhostRegex.test(href) || mailtoRegex.test(href);
+    }
+    // Inspect non-navigation links and adorn them if external
+ 	var links = window.document.querySelectorAll('a[href]:not(.nav-link):not(.navbar-brand):not(.toc-action):not(.sidebar-link):not(.sidebar-item-toggle):not(.pagination-link):not(.no-external):not([aria-hidden]):not(.dropdown-item):not(.quarto-navigation-tool)');
+    for (var i=0; i<links.length; i++) {
+      const link = links[i];
+      if (!isInternal(link.href)) {
+        // undo the damage that might have been done by quarto-nav.js in the case of
+        // links that we want to consider external
+        if (link.dataset.originalHref !== undefined) {
+          link.href = link.dataset.originalHref;
+        }
+      }
+    }
+  function tippyHover(el, contentFn, onTriggerFn, onUntriggerFn) {
+    const config = {
+      allowHTML: true,
+      maxWidth: 500,
+      delay: 100,
+      arrow: false,
+      appendTo: function(el) {
+          return el.parentElement;
+      },
+      interactive: true,
+      interactiveBorder: 10,
+      theme: 'quarto',
+      placement: 'bottom-start',
+    };
+    if (contentFn) {
+      config.content = contentFn;
+    }
+    if (onTriggerFn) {
+      config.onTrigger = onTriggerFn;
+    }
+    if (onUntriggerFn) {
+      config.onUntrigger = onUntriggerFn;
+    }
+    window.tippy(el, config); 
+  }
+  const noterefs = window.document.querySelectorAll('a[role="doc-noteref"]');
+  for (var i=0; i<noterefs.length; i++) {
+    const ref = noterefs[i];
+    tippyHover(ref, function() {
+      // use id or data attribute instead here
+      let href = ref.getAttribute('data-footnote-href') || ref.getAttribute('href');
+      try { href = new URL(href).hash; } catch {}
+      const id = href.replace(/^#\/?/, "");
+      const note = window.document.getElementById(id);
+      if (note) {
+        return note.innerHTML;
+      } else {
+        return "";
+      }
+    });
+  }
+  const xrefs = window.document.querySelectorAll('a.quarto-xref');
+  const processXRef = (id, note) => {
+    // Strip column container classes
+    const stripColumnClz = (el) => {
+      el.classList.remove("page-full", "page-columns");
+      if (el.children) {
+        for (const child of el.children) {
+          stripColumnClz(child);
+        }
+      }
+    }
+    stripColumnClz(note)
+    if (id === null || id.startsWith('sec-')) {
+      // Special case sections, only their first couple elements
+      const container = document.createElement("div");
+      if (note.children && note.children.length > 2) {
+        container.appendChild(note.children[0].cloneNode(true));
+        for (let i = 1; i < note.children.length; i++) {
+          const child = note.children[i];
+          if (child.tagName === "P" && child.innerText === "") {
+            continue;
+          } else {
+            container.appendChild(child.cloneNode(true));
+            break;
+          }
+        }
+        if (window.Quarto?.typesetMath) {
+          window.Quarto.typesetMath(container);
+        }
+        return container.innerHTML
+      } else {
+        if (window.Quarto?.typesetMath) {
+          window.Quarto.typesetMath(note);
+        }
+        return note.innerHTML;
+      }
+    } else {
+      // Remove any anchor links if they are present
+      const anchorLink = note.querySelector('a.anchorjs-link');
+      if (anchorLink) {
+        anchorLink.remove();
+      }
+      if (window.Quarto?.typesetMath) {
+        window.Quarto.typesetMath(note);
+      }
+      // TODO in 1.5, we should make sure this works without a callout special case
+      if (note.classList.contains("callout")) {
+        return note.outerHTML;
+      } else {
+        return note.innerHTML;
+      }
+    }
+  }
+  for (var i=0; i<xrefs.length; i++) {
+    const xref = xrefs[i];
+    tippyHover(xref, undefined, function(instance) {
+      instance.disable();
+      let url = xref.getAttribute('href');
+      let hash = undefined; 
+      if (url.startsWith('#')) {
+        hash = url;
+      } else {
+        try { hash = new URL(url).hash; } catch {}
+      }
+      if (hash) {
+        const id = hash.replace(/^#\/?/, "");
+        const note = window.document.getElementById(id);
+        if (note !== null) {
+          try {
+            const html = processXRef(id, note.cloneNode(true));
+            instance.setContent(html);
+          } finally {
+            instance.enable();
+            instance.show();
+          }
+        } else {
+          // See if we can fetch this
+          fetch(url.split('#')[0])
+          .then(res => res.text())
+          .then(html => {
+            const parser = new DOMParser();
+            const htmlDoc = parser.parseFromString(html, "text/html");
+            const note = htmlDoc.getElementById(id);
+            if (note !== null) {
+              const html = processXRef(id, note);
+              instance.setContent(html);
+            } 
+          }).finally(() => {
+            instance.enable();
+            instance.show();
+          });
+        }
+      } else {
+        // See if we can fetch a full url (with no hash to target)
+        // This is a special case and we should probably do some content thinning / targeting
+        fetch(url)
+        .then(res => res.text())
+        .then(html => {
+          const parser = new DOMParser();
+          const htmlDoc = parser.parseFromString(html, "text/html");
+          const note = htmlDoc.querySelector('main.content');
+          if (note !== null) {
+            // This should only happen for chapter cross references
+            // (since there is no id in the URL)
+            // remove the first header
+            if (note.children.length > 0 && note.children[0].tagName === "HEADER") {
+              note.children[0].remove();
+            }
+            const html = processXRef(null, note);
+            instance.setContent(html);
+          } 
+        }).finally(() => {
+          instance.enable();
+          instance.show();
+        });
+      }
+    }, function(instance) {
+    });
+  }
+      let selectedAnnoteEl;
+      const selectorForAnnotation = ( cell, annotation) => {
+        let cellAttr = 'data-code-cell="' + cell + '"';
+        let lineAttr = 'data-code-annotation="' +  annotation + '"';
+        const selector = 'span[' + cellAttr + '][' + lineAttr + ']';
+        return selector;
+      }
+      const selectCodeLines = (annoteEl) => {
+        const doc = window.document;
+        const targetCell = annoteEl.getAttribute("data-target-cell");
+        const targetAnnotation = annoteEl.getAttribute("data-target-annotation");
+        const annoteSpan = window.document.querySelector(selectorForAnnotation(targetCell, targetAnnotation));
+        const lines = annoteSpan.getAttribute("data-code-lines").split(",");
+        const lineIds = lines.map((line) => {
+          return targetCell + "-" + line;
+        })
+        let top = null;
+        let height = null;
+        let parent = null;
+        if (lineIds.length > 0) {
+            //compute the position of the single el (top and bottom and make a div)
+            const el = window.document.getElementById(lineIds[0]);
+            top = el.offsetTop;
+            height = el.offsetHeight;
+            parent = el.parentElement.parentElement;
+          if (lineIds.length > 1) {
+            const lastEl = window.document.getElementById(lineIds[lineIds.length - 1]);
+            const bottom = lastEl.offsetTop + lastEl.offsetHeight;
+            height = bottom - top;
+          }
+          if (top !== null && height !== null && parent !== null) {
+            // cook up a div (if necessary) and position it 
+            let div = window.document.getElementById("code-annotation-line-highlight");
+            if (div === null) {
+              div = window.document.createElement("div");
+              div.setAttribute("id", "code-annotation-line-highlight");
+              div.style.position = 'absolute';
+              parent.appendChild(div);
+            }
+            div.style.top = top - 2 + "px";
+            div.style.height = height + 4 + "px";
+            div.style.left = 0;
+            let gutterDiv = window.document.getElementById("code-annotation-line-highlight-gutter");
+            if (gutterDiv === null) {
+              gutterDiv = window.document.createElement("div");
+              gutterDiv.setAttribute("id", "code-annotation-line-highlight-gutter");
+              gutterDiv.style.position = 'absolute';
+              const codeCell = window.document.getElementById(targetCell);
+              const gutter = codeCell.querySelector('.code-annotation-gutter');
+              gutter.appendChild(gutterDiv);
+            }
+            gutterDiv.style.top = top - 2 + "px";
+            gutterDiv.style.height = height + 4 + "px";
+          }
+          selectedAnnoteEl = annoteEl;
+        }
+      };
+      const unselectCodeLines = () => {
+        const elementsIds = ["code-annotation-line-highlight", "code-annotation-line-highlight-gutter"];
+        elementsIds.forEach((elId) => {
+          const div = window.document.getElementById(elId);
+          if (div) {
+            div.remove();
+          }
+        });
+        selectedAnnoteEl = undefined;
+      };
+        // Handle positioning of the toggle
+    window.addEventListener(
+      "resize",
+      throttle(() => {
+        elRect = undefined;
+        if (selectedAnnoteEl) {
+          selectCodeLines(selectedAnnoteEl);
+        }
+      }, 10)
+    );
+    function throttle(fn, ms) {
+    let throttle = false;
+    let timer;
+      return (...args) => {
+        if(!throttle) { // first call gets through
+            fn.apply(this, args);
+            throttle = true;
+        } else { // all the others get throttled
+            if(timer) clearTimeout(timer); // cancel #2
+            timer = setTimeout(() => {
+              fn.apply(this, args);
+              timer = throttle = false;
+            }, ms);
+        }
+      };
+    }
+      // Attach click handler to the DT
+      const annoteDls = window.document.querySelectorAll('dt[data-target-cell]');
+      for (const annoteDlNode of annoteDls) {
+        annoteDlNode.addEventListener('click', (event) => {
+          const clickedEl = event.target;
+          if (clickedEl !== selectedAnnoteEl) {
+            unselectCodeLines();
+            const activeEl = window.document.querySelector('dt[data-target-cell].code-annotation-active');
+            if (activeEl) {
+              activeEl.classList.remove('code-annotation-active');
+            }
+            selectCodeLines(clickedEl);
+            clickedEl.classList.add('code-annotation-active');
+          } else {
+            // Unselect the line
+            unselectCodeLines();
+            clickedEl.classList.remove('code-annotation-active');
+          }
+        });
+      }
+  const findCites = (el) => {
+    const parentEl = el.parentElement;
+    if (parentEl) {
+      const cites = parentEl.dataset.cites;
+      if (cites) {
+        return {
+          el,
+          cites: cites.split(' ')
+        };
+      } else {
+        return findCites(el.parentElement)
+      }
+    } else {
+      return undefined;
+    }
+  };
+  var bibliorefs = window.document.querySelectorAll('a[role="doc-biblioref"]');
+  for (var i=0; i<bibliorefs.length; i++) {
+    const ref = bibliorefs[i];
+    const citeInfo = findCites(ref);
+    if (citeInfo) {
+      tippyHover(citeInfo.el, function() {
+        var popup = window.document.createElement('div');
+        citeInfo.cites.forEach(function(cite) {
+          var citeDiv = window.document.createElement('div');
+          citeDiv.classList.add('hanging-indent');
+          citeDiv.classList.add('csl-entry');
+          var biblioDiv = window.document.getElementById('ref-' + cite);
+          if (biblioDiv) {
+            citeDiv.innerHTML = biblioDiv.innerHTML;
+          }
+          popup.appendChild(citeDiv);
+        });
+        return popup.innerHTML;
+      });
+    }
+  }
+});
+</script>
+</div> <!-- /content -->
+
+
+
+
+</body></html>
\ No newline at end of file
diff --git a/docs/config.html b/docs/config.html
index 93df908c3..bdb9020be 100644
--- a/docs/config.html
+++ b/docs/config.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/dataset-formats/conversation.html b/docs/dataset-formats/conversation.html
index e326976f9..78db31170 100644
--- a/docs/dataset-formats/conversation.html
+++ b/docs/dataset-formats/conversation.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text">Conversation</span></a>
+  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link active">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -281,15 +281,11 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
     <h2 id="toc-title">On this page</h2>
    
   <ul>
-  <li><a href="#formats" id="toc-formats" class="nav-link active" data-scroll-target="#formats">Formats</a>
-  <ul class="collapse">
-  <li><a href="#sharegpt" id="toc-sharegpt" class="nav-link" data-scroll-target="#sharegpt">sharegpt</a></li>
+  <li><a href="#sharegpt" id="toc-sharegpt" class="nav-link active" data-scroll-target="#sharegpt">sharegpt</a></li>
   <li><a href="#pygmalion" id="toc-pygmalion" class="nav-link" data-scroll-target="#pygmalion">pygmalion</a></li>
   <li><a href="#sharegpt.load_role" id="toc-sharegpt.load_role" class="nav-link" data-scroll-target="#sharegpt.load_role">sharegpt.load_role</a></li>
   <li><a href="#sharegpt.load_guanaco" id="toc-sharegpt.load_guanaco" class="nav-link" data-scroll-target="#sharegpt.load_guanaco">sharegpt.load_guanaco</a></li>
   <li><a href="#sharegpt_jokes" id="toc-sharegpt_jokes" class="nav-link" data-scroll-target="#sharegpt_jokes">sharegpt_jokes</a></li>
-  </ul></li>
-  <li><a href="#how-to-add-custom-prompts-for-instruction-tuning" id="toc-how-to-add-custom-prompts-for-instruction-tuning" class="nav-link" data-scroll-target="#how-to-add-custom-prompts-for-instruction-tuning">How to add custom prompts for instruction-tuning</a></li>
   </ul>
 </nav>
     </div>
@@ -320,10 +316,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </header>
 
 
-<section id="formats" class="level2">
-<h2 class="anchored" data-anchor-id="formats">Formats</h2>
-<section id="sharegpt" class="level3">
-<h3 class="anchored" data-anchor-id="sharegpt">sharegpt</h3>
+<section id="sharegpt" class="level2">
+<h2 class="anchored" data-anchor-id="sharegpt">sharegpt</h2>
 <p>conversations where <code>from</code> is <code>human</code>/<code>gpt</code>. (optional: first row with role <code>system</code> to override default system prompt)</p>
 <div class="code-with-filename">
 <div class="code-with-filename-file">
@@ -331,20 +325,21 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </div>
 <div class="sourceCode" id="cb1" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb1-1"><a href="#cb1-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"from"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"value"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </div>
-<p>Note: <code>type: sharegpt</code> opens a special config <code>conversation:</code> that enables conversions to many Conversation types. See <a href="../docs/config.qmd">the docs</a> for all config options.</p>
+<p>Note: <code>type: sharegpt</code> opens special configs: - <code>conversation</code>: enables conversions to many Conversation types. Refer to the ‘name’ <a href="https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py">here</a> for options. - <code>roles</code>: allows you to specify the roles for input and output. This is useful for datasets with custom roles such as <code>tool</code> etc to support masking. - <code>field_human</code>: specify the key to use instead of <code>human</code> in the conversation. - <code>field_model</code>: specify the key to use instead of <code>gpt</code> in the conversation.</p>
+<div class="sourceCode" id="cb2"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb2-1"><a href="#cb2-1" aria-hidden="true" tabindex="-1"></a><span class="fu">datasets</span><span class="kw">:</span></span>
+<span id="cb2-2"><a href="#cb2-2" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">path</span><span class="kw">:</span><span class="at"> ...</span></span>
+<span id="cb2-3"><a href="#cb2-3" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> sharegpt</span></span>
+<span id="cb2-4"><a href="#cb2-4" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb2-5"><a href="#cb2-5" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">conversation</span><span class="kw">:</span><span class="co"> # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py</span></span>
+<span id="cb2-6"><a href="#cb2-6" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">field_human</span><span class="kw">:</span><span class="co"> # Optional[str]. Human key to use for conversation.</span></span>
+<span id="cb2-7"><a href="#cb2-7" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">field_model</span><span class="kw">:</span><span class="co"> # Optional[str]. Assistant key to use for conversation.</span></span>
+<span id="cb2-8"><a href="#cb2-8" aria-hidden="true" tabindex="-1"></a><span class="co">    # Add additional keys from your dataset as input or output roles</span></span>
+<span id="cb2-9"><a href="#cb2-9" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">roles</span><span class="kw">:</span></span>
+<span id="cb2-10"><a href="#cb2-10" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">input</span><span class="kw">:</span><span class="co"> # Optional[List[str]]. These will be masked based on train_on_input</span></span>
+<span id="cb2-11"><a href="#cb2-11" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">output</span><span class="kw">:</span><span class="co"> # Optional[List[str]].</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </section>
-<section id="pygmalion" class="level3">
-<h3 class="anchored" data-anchor-id="pygmalion">pygmalion</h3>
-<div class="code-with-filename">
-<div class="code-with-filename-file">
-<pre><strong>data.jsonl</strong></pre>
-</div>
-<div class="sourceCode" id="cb2" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb2-1"><a href="#cb2-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"role"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"value"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
-</div>
-</section>
-<section id="sharegpt.load_role" class="level3">
-<h3 class="anchored" data-anchor-id="sharegpt.load_role">sharegpt.load_role</h3>
-<p>conversations where <code>role</code> is used instead of <code>from</code></p>
+<section id="pygmalion" class="level2">
+<h2 class="anchored" data-anchor-id="pygmalion">pygmalion</h2>
 <div class="code-with-filename">
 <div class="code-with-filename-file">
 <pre><strong>data.jsonl</strong></pre>
@@ -352,52 +347,35 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 <div class="sourceCode" id="cb3" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb3-1"><a href="#cb3-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"role"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"value"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </div>
 </section>
-<section id="sharegpt.load_guanaco" class="level3">
-<h3 class="anchored" data-anchor-id="sharegpt.load_guanaco">sharegpt.load_guanaco</h3>
+<section id="sharegpt.load_role" class="level2">
+<h2 class="anchored" data-anchor-id="sharegpt.load_role">sharegpt.load_role</h2>
+<p>conversations where <code>role</code> is used instead of <code>from</code></p>
+<div class="code-with-filename">
+<div class="code-with-filename-file">
+<pre><strong>data.jsonl</strong></pre>
+</div>
+<div class="sourceCode" id="cb4" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb4-1"><a href="#cb4-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"role"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"value"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+</div>
+</section>
+<section id="sharegpt.load_guanaco" class="level2">
+<h2 class="anchored" data-anchor-id="sharegpt.load_guanaco">sharegpt.load_guanaco</h2>
 <p>conversations where <code>from</code> is <code>prompter</code> <code>assistant</code> instead of default sharegpt</p>
 <div class="code-with-filename">
 <div class="code-with-filename-file">
 <pre><strong>data.jsonl</strong></pre>
 </div>
-<div class="sourceCode" id="cb4" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb4-1"><a href="#cb4-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"from"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"value"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb5" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb5-1"><a href="#cb5-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"from"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"value"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </div>
 </section>
-<section id="sharegpt_jokes" class="level3">
-<h3 class="anchored" data-anchor-id="sharegpt_jokes">sharegpt_jokes</h3>
+<section id="sharegpt_jokes" class="level2">
+<h2 class="anchored" data-anchor-id="sharegpt_jokes">sharegpt_jokes</h2>
 <p>creates a chat where bot is asked to tell a joke, then explain why the joke is funny</p>
 <div class="code-with-filename">
 <div class="code-with-filename-file">
 <pre><strong>data.jsonl</strong></pre>
 </div>
-<div class="sourceCode" id="cb5" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb5-1"><a href="#cb5-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"title"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"text"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"explanation"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb6" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb6-1"><a href="#cb6-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"conversations"</span><span class="fu">:</span> <span class="ot">[</span><span class="fu">{</span><span class="dt">"title"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"text"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"explanation"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span><span class="ot">]</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </div>
-</section>
-</section>
-<section id="how-to-add-custom-prompts-for-instruction-tuning" class="level2">
-<h2 class="anchored" data-anchor-id="how-to-add-custom-prompts-for-instruction-tuning">How to add custom prompts for instruction-tuning</h2>
-<p>For a dataset that is preprocessed for instruction purposes:</p>
-<div class="code-with-filename">
-<div class="code-with-filename-file">
-<pre><strong>data.jsonl</strong></pre>
-</div>
-<div class="sourceCode" id="cb6" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb6-1"><a href="#cb6-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"input"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"output"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
-</div>
-<p>You can use this example in your YAML config:</p>
-<div class="code-with-filename">
-<div class="code-with-filename-file">
-<pre><strong>config.yaml</strong></pre>
-</div>
-<div class="sourceCode" id="cb7" data-filename="config.yaml"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb7-1"><a href="#cb7-1" aria-hidden="true" tabindex="-1"></a><span class="fu">datasets</span><span class="kw">:</span></span>
-<span id="cb7-2"><a href="#cb7-2" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> repo</span></span>
-<span id="cb7-3"><a href="#cb7-3" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span></span>
-<span id="cb7-4"><a href="#cb7-4" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">system_prompt</span><span class="kw">:</span><span class="at"> </span><span class="st">""</span></span>
-<span id="cb7-5"><a href="#cb7-5" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">field_system</span><span class="kw">:</span><span class="at"> system</span></span>
-<span id="cb7-6"><a href="#cb7-6" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">field_instruction</span><span class="kw">:</span><span class="at"> input</span></span>
-<span id="cb7-7"><a href="#cb7-7" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">field_output</span><span class="kw">:</span><span class="at"> output</span></span>
-<span id="cb7-8"><a href="#cb7-8" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">format</span><span class="kw">:</span><span class="at"> </span><span class="st">"[INST] {instruction} [/INST]"</span></span>
-<span id="cb7-9"><a href="#cb7-9" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">no_input_format</span><span class="kw">:</span><span class="at"> </span><span class="st">"[INST] {instruction} [/INST]"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
-</div>
-<p>See full config options under <a href="../docs/config.qmd">here</a>.</p>
 
 
 </section>
diff --git a/docs/dataset-formats/index.html b/docs/dataset-formats/index.html
index bc7a8ef0d..a28275d93 100644
--- a/docs/dataset-formats/index.html
+++ b/docs/dataset-formats/index.html
@@ -244,8 +244,8 @@ window.Quarto = {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -256,8 +256,8 @@ window.Quarto = {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -351,23 +351,7 @@ Description
 </tr>
 </thead>
 <tbody class="list">
-<tr data-index="0" data-listing-file-modified-sort="1712171163568" data-listing-reading-time-sort="1" data-listing-word-count-sort="189" data-listing-title-sort="Conversation" data-listing-filename-sort="conversation.qmd">
-<td>
-<a href="../../docs/dataset-formats/conversation.html" class="title listing-title">Conversation</a>
-</td>
-<td>
-<span class="listing-description">Conversation format for supervised fine-tuning.</span>
-</td>
-</tr>
-<tr data-index="1" data-listing-file-modified-sort="1712171163568" data-listing-reading-time-sort="2" data-listing-word-count-sort="222" data-listing-title-sort="Instruction Tuning" data-listing-filename-sort="inst_tune.qmd">
-<td>
-<a href="../../docs/dataset-formats/inst_tune.html" class="title listing-title">Instruction Tuning</a>
-</td>
-<td>
-<span class="listing-description">Instruction tuning formats for supervised fine-tuning.</span>
-</td>
-</tr>
-<tr data-index="2" data-listing-file-modified-sort="1712171163568" data-listing-reading-time-sort="1" data-listing-word-count-sort="47" data-listing-title-sort="Pre-training" data-listing-filename-sort="pretraining.qmd">
+<tr data-index="0" data-listing-file-modified-sort="1712273633798" data-listing-reading-time-sort="1" data-listing-word-count-sort="47" data-listing-title-sort="Pre-training" data-listing-filename-sort="pretraining.qmd">
 <td>
 <a href="../../docs/dataset-formats/pretraining.html" class="title listing-title">Pre-training</a>
 </td>
@@ -375,7 +359,23 @@ Description
 <span class="listing-description">Data format for a pre-training completion task.</span>
 </td>
 </tr>
-<tr data-index="3" data-listing-file-modified-sort="1712171163568" data-listing-reading-time-sort="1" data-listing-word-count-sort="3" data-listing-title-sort="Template-Free" data-listing-filename-sort="template_free.qmd">
+<tr data-index="1" data-listing-file-modified-sort="1712273633798" data-listing-reading-time-sort="2" data-listing-word-count-sort="308" data-listing-title-sort="Instruction Tuning" data-listing-filename-sort="inst_tune.qmd">
+<td>
+<a href="../../docs/dataset-formats/inst_tune.html" class="title listing-title">Instruction Tuning</a>
+</td>
+<td>
+<span class="listing-description">Instruction tuning formats for supervised fine-tuning.</span>
+</td>
+</tr>
+<tr data-index="2" data-listing-file-modified-sort="1712273633798" data-listing-reading-time-sort="2" data-listing-word-count-sort="235" data-listing-title-sort="Conversation" data-listing-filename-sort="conversation.qmd">
+<td>
+<a href="../../docs/dataset-formats/conversation.html" class="title listing-title">Conversation</a>
+</td>
+<td>
+<span class="listing-description">Conversation format for supervised fine-tuning.</span>
+</td>
+</tr>
+<tr data-index="3" data-listing-file-modified-sort="1712273633798" data-listing-reading-time-sort="1" data-listing-word-count-sort="3" data-listing-title-sort="Template-Free" data-listing-filename-sort="template_free.qmd">
 <td>
 <a href="../../docs/dataset-formats/template_free.html" class="title listing-title">Template-Free</a>
 </td>
@@ -383,7 +383,7 @@ Description
 <span class="listing-description">Construct prompts without a template.</span>
 </td>
 </tr>
-<tr data-index="4" data-listing-file-modified-sort="1712171163568" data-listing-reading-time-sort="1" data-listing-word-count-sort="22" data-listing-title-sort="Custom Pre-Tokenized Dataset" data-listing-filename-sort="tokenized.qmd">
+<tr data-index="4" data-listing-file-modified-sort="1712273633798" data-listing-reading-time-sort="1" data-listing-word-count-sort="22" data-listing-title-sort="Custom Pre-Tokenized Dataset" data-listing-filename-sort="tokenized.qmd">
 <td>
 <a href="../../docs/dataset-formats/tokenized.html" class="title listing-title">Custom Pre-Tokenized Dataset</a>
 </td>
diff --git a/docs/dataset-formats/inst_tune.html b/docs/dataset-formats/inst_tune.html
index 0a9735cd8..88455cb5b 100644
--- a/docs/dataset-formats/inst_tune.html
+++ b/docs/dataset-formats/inst_tune.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -301,6 +301,7 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
   <li><a href="#creative_acr.load_critique" id="toc-creative_acr.load_critique" class="nav-link" data-scroll-target="#creative_acr.load_critique">creative_acr.load_critique</a></li>
   <li><a href="#creative_acr.load_revise" id="toc-creative_acr.load_revise" class="nav-link" data-scroll-target="#creative_acr.load_revise">creative_acr.load_revise</a></li>
   <li><a href="#metharme" id="toc-metharme" class="nav-link" data-scroll-target="#metharme">metharme</a></li>
+  <li><a href="#how-to-add-custom-prompt-format" id="toc-how-to-add-custom-prompt-format" class="nav-link" data-scroll-target="#how-to-add-custom-prompt-format">How to add custom prompt format</a></li>
   </ul>
 </nav>
     </div>
@@ -530,6 +531,32 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </div>
 <div class="sourceCode" id="cb20" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb20-1"><a href="#cb20-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"prompt"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"generation"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </div>
+</section>
+<section id="how-to-add-custom-prompt-format" class="level2">
+<h2 class="anchored" data-anchor-id="how-to-add-custom-prompt-format">How to add custom prompt format</h2>
+<p>For a dataset that is preprocessed for instruction purposes:</p>
+<div class="code-with-filename">
+<div class="code-with-filename-file">
+<pre><strong>data.jsonl</strong></pre>
+</div>
+<div class="sourceCode" id="cb21" data-filename="data.jsonl"><pre class="sourceCode json code-with-copy"><code class="sourceCode json"><span id="cb21-1"><a href="#cb21-1" aria-hidden="true" tabindex="-1"></a><span class="fu">{</span><span class="dt">"input"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">,</span> <span class="dt">"output"</span><span class="fu">:</span> <span class="st">"..."</span><span class="fu">}</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+</div>
+<p>You can use this example in your YAML config:</p>
+<div class="code-with-filename">
+<div class="code-with-filename-file">
+<pre><strong>config.yaml</strong></pre>
+</div>
+<div class="sourceCode" id="cb22" data-filename="config.yaml"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb22-1"><a href="#cb22-1" aria-hidden="true" tabindex="-1"></a><span class="fu">datasets</span><span class="kw">:</span></span>
+<span id="cb22-2"><a href="#cb22-2" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> repo</span></span>
+<span id="cb22-3"><a href="#cb22-3" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span></span>
+<span id="cb22-4"><a href="#cb22-4" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">system_prompt</span><span class="kw">:</span><span class="at"> </span><span class="st">""</span></span>
+<span id="cb22-5"><a href="#cb22-5" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">field_system</span><span class="kw">:</span><span class="at"> system</span></span>
+<span id="cb22-6"><a href="#cb22-6" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">field_instruction</span><span class="kw">:</span><span class="at"> input</span></span>
+<span id="cb22-7"><a href="#cb22-7" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">field_output</span><span class="kw">:</span><span class="at"> output</span></span>
+<span id="cb22-8"><a href="#cb22-8" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">format</span><span class="kw">:</span><span class="at"> </span><span class="st">"[INST] {instruction} [/INST]"</span></span>
+<span id="cb22-9"><a href="#cb22-9" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="fu">no_input_format</span><span class="kw">:</span><span class="at"> </span><span class="st">"[INST] {instruction} [/INST]"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+</div>
+<p>See full config options under <a href="../../docs/config.html">here</a>.</p>
 
 
 </section>
diff --git a/docs/dataset-formats/pretraining.html b/docs/dataset-formats/pretraining.html
index 1bd48f7ba..1ec65cdcc 100644
--- a/docs/dataset-formats/pretraining.html
+++ b/docs/dataset-formats/pretraining.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link active">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link active">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/dataset-formats/template_free.html b/docs/dataset-formats/template_free.html
index 23edcc054..20a3a131b 100644
--- a/docs/dataset-formats/template_free.html
+++ b/docs/dataset-formats/template_free.html
@@ -184,8 +184,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -196,8 +196,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/dataset-formats/tokenized.html b/docs/dataset-formats/tokenized.html
index 9587f0df8..c66b037d2 100644
--- a/docs/dataset-formats/tokenized.html
+++ b/docs/dataset-formats/tokenized.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/debugging.html b/docs/debugging.html
index f68f917a1..b07ef0695 100644
--- a/docs/debugging.html
+++ b/docs/debugging.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/faq.html b/docs/faq.html
index 9882cd611..9da1b4d8f 100644
--- a/docs/faq.html
+++ b/docs/faq.html
@@ -184,8 +184,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -196,8 +196,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/fsdp_qlora.html b/docs/fsdp_qlora.html
index e1f6f3f77..d761e0860 100644
--- a/docs/fsdp_qlora.html
+++ b/docs/fsdp_qlora.html
@@ -184,8 +184,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -196,8 +196,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/input_output.html b/docs/input_output.html
index f39d3be67..fd054a7ac 100644
--- a/docs/input_output.html
+++ b/docs/input_output.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -352,7 +352,7 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </section>
 <section id="you-may-not-want-prompt-templates" class="level3">
 <h3 class="anchored" data-anchor-id="you-may-not-want-prompt-templates">You may not want prompt templates</h3>
-<p>However, there are many situations where you don’t want to use one of these formats or templates (I usually don’t!). This is because they can:</p>
+<p>However, there are many situations where you don’t want to use one of these formats or templates. This is because they can:</p>
 <ul>
 <li>Add unnecessary boilerplate to your prompts.</li>
 <li>Create artifacts like special delimiters <code>&lt;|im_start|&gt;</code> that can quickly become footguns if you don’t include them correctly at inference time.</li>
diff --git a/docs/mac.html b/docs/mac.html
index e2b0de21e..0c71e59a2 100644
--- a/docs/mac.html
+++ b/docs/mac.html
@@ -184,8 +184,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -196,8 +196,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/multi-node.html b/docs/multi-node.html
index 56e36518d..cfd539edd 100644
--- a/docs/multi-node.html
+++ b/docs/multi-node.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/multipack.html b/docs/multipack.html
index dd70593bd..7fe8333ee 100644
--- a/docs/multipack.html
+++ b/docs/multipack.html
@@ -184,8 +184,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -196,8 +196,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/nccl.html b/docs/nccl.html
index 33bfb8c16..6d158b646 100644
--- a/docs/nccl.html
+++ b/docs/nccl.html
@@ -184,8 +184,8 @@ ul.task-list li input[type="checkbox"] {
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -196,8 +196,8 @@ ul.task-list li input[type="checkbox"] {
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/docs/rlhf.html b/docs/rlhf.html
index 2bc0bba04..b9e235697 100644
--- a/docs/rlhf.html
+++ b/docs/rlhf.html
@@ -218,8 +218,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -230,8 +230,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/examples/colab-notebooks/colab-axolotl-example.html b/examples/colab-notebooks/colab-axolotl-example.html
index b7fb503bd..2bbbef8a1 100644
--- a/examples/colab-notebooks/colab-axolotl-example.html
+++ b/examples/colab-notebooks/colab-axolotl-example.html
@@ -217,8 +217,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -229,8 +229,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="../../docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="../../docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
diff --git a/index.html b/index.html
index 2c492972b..788bac38b 100644
--- a/index.html
+++ b/index.html
@@ -217,8 +217,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
       <ul id="quarto-sidebar-section-2" class="collapse list-unstyled sidebar-section depth1 show">  
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="./docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Conversation</span></a>
+  <a href="./docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Pre-training</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -229,8 +229,8 @@ pre > code.sourceCode > span > a:first-child::before { text-decoration: underlin
 </li>
           <li class="sidebar-item">
   <div class="sidebar-item-container"> 
-  <a href="./docs/dataset-formats/pretraining.html" class="sidebar-item-text sidebar-link">
- <span class="menu-text">Pre-training</span></a>
+  <a href="./docs/dataset-formats/conversation.html" class="sidebar-item-text sidebar-link">
+ <span class="menu-text">Conversation</span></a>
   </div>
 </li>
           <li class="sidebar-item">
@@ -616,23 +616,16 @@ Click to Expand
 <div class="sourceCode" id="cb7"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb7-1"><a href="#cb7-1" aria-hidden="true" tabindex="-1"></a><span class="fu">wget</span> https://bootstrap.pypa.io/get-pip.py</span>
 <span id="cb7-2"><a href="#cb7-2" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> get-pip.py</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <ol start="3" type="1">
-<li>Install torch</li>
+<li><p>Install Pytorch https://pytorch.org/get-started/locally/</p></li>
+<li><p>Follow instructions on quickstart.</p></li>
+<li><p>Run</p></li>
 </ol>
-<div class="sourceCode" id="cb8"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb8-1"><a href="#cb8-1" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install <span class="at">-U</span> torch <span class="at">--index-url</span> https://download.pytorch.org/whl/cu118</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
-<ol start="4" type="1">
-<li>Axolotl</li>
-</ol>
-<div class="sourceCode" id="cb9"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb9-1"><a href="#cb9-1" aria-hidden="true" tabindex="-1"></a><span class="fu">git</span> clone https://github.com/OpenAccess-AI-Collective/axolotl</span>
-<span id="cb9-2"><a href="#cb9-2" aria-hidden="true" tabindex="-1"></a><span class="bu">cd</span> axolotl</span>
-<span id="cb9-3"><a href="#cb9-3" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb9-4"><a href="#cb9-4" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install packaging</span>
-<span id="cb9-5"><a href="#cb9-5" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install <span class="at">-e</span> <span class="st">'.[flash-attn,deepspeed]'</span></span>
-<span id="cb9-6"><a href="#cb9-6" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install protobuf==3.20.3</span>
-<span id="cb9-7"><a href="#cb9-7" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install <span class="at">-U</span> <span class="at">--ignore-installed</span> requests Pillow psutil scipy</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
-<ol start="5" type="1">
+<div class="sourceCode" id="cb8"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb8-1"><a href="#cb8-1" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install protobuf==3.20.3</span>
+<span id="cb8-2"><a href="#cb8-2" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install <span class="at">-U</span> <span class="at">--ignore-installed</span> requests Pillow psutil scipy</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<ol start="6" type="1">
 <li>Set path</li>
 </ol>
-<div class="sourceCode" id="cb10"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb10-1"><a href="#cb10-1" aria-hidden="true" tabindex="-1"></a><span class="bu">export</span> <span class="va">LD_LIBRARY_PATH</span><span class="op">=</span>/usr/lib/x86_64-linux-gnu:<span class="va">$LD_LIBRARY_PATH</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb9"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb9-1"><a href="#cb9-1" aria-hidden="true" tabindex="-1"></a><span class="bu">export</span> <span class="va">LD_LIBRARY_PATH</span><span class="op">=</span>/usr/lib/x86_64-linux-gnu:<span class="va">$LD_LIBRARY_PATH</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </details>
 </section>
 <section id="gcp" class="level5">
@@ -643,7 +636,7 @@ Click to Expand
 </summary>
 <p>Use a Deeplearning linux OS with cuda and pytorch installed. Then follow instructions on quickstart.</p>
 <p>Make sure to run the below to uninstall xla.</p>
-<div class="sourceCode" id="cb11"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb11-1"><a href="#cb11-1" aria-hidden="true" tabindex="-1"></a><span class="ex">pip</span> uninstall <span class="at">-y</span> torch_xla<span class="pp">[</span><span class="ss">tpu</span><span class="pp">]</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb10"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb10-1"><a href="#cb10-1" aria-hidden="true" tabindex="-1"></a><span class="ex">pip</span> uninstall <span class="at">-y</span> torch_xla<span class="pp">[</span><span class="ss">tpu</span><span class="pp">]</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </details>
 </section>
 </section>
@@ -664,17 +657,17 @@ Click to Expand
 <section id="launching-on-public-clouds-via-skypilot" class="level4">
 <h4 class="anchored" data-anchor-id="launching-on-public-clouds-via-skypilot">Launching on public clouds via SkyPilot</h4>
 <p>To launch on GPU instances (both on-demand and spot instances) on 7+ clouds (GCP, AWS, Azure, OCI, and more), you can use <a href="https://skypilot.readthedocs.io/en/latest/index.html">SkyPilot</a>:</p>
-<div class="sourceCode" id="cb13"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb13-1"><a href="#cb13-1" aria-hidden="true" tabindex="-1"></a><span class="ex">pip</span> install <span class="st">"skypilot-nightly[gcp,aws,azure,oci,lambda,kubernetes,ibm,scp]"</span>  <span class="co"># choose your clouds</span></span>
-<span id="cb13-2"><a href="#cb13-2" aria-hidden="true" tabindex="-1"></a><span class="ex">sky</span> check</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb12"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb12-1"><a href="#cb12-1" aria-hidden="true" tabindex="-1"></a><span class="ex">pip</span> install <span class="st">"skypilot-nightly[gcp,aws,azure,oci,lambda,kubernetes,ibm,scp]"</span>  <span class="co"># choose your clouds</span></span>
+<span id="cb12-2"><a href="#cb12-2" aria-hidden="true" tabindex="-1"></a><span class="ex">sky</span> check</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>Get the <a href="https://github.com/skypilot-org/skypilot/tree/master/llm/axolotl">example YAMLs</a> of using Axolotl to finetune <code>mistralai/Mistral-7B-v0.1</code>:</p>
 <pre><code>git clone https://github.com/skypilot-org/skypilot.git
 cd skypilot/llm/axolotl</code></pre>
 <p>Use one command to launch:</p>
-<div class="sourceCode" id="cb15"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb15-1"><a href="#cb15-1" aria-hidden="true" tabindex="-1"></a><span class="co"># On-demand</span></span>
-<span id="cb15-2"><a href="#cb15-2" aria-hidden="true" tabindex="-1"></a><span class="va">HF_TOKEN</span><span class="op">=</span>xx <span class="ex">sky</span> launch axolotl.yaml <span class="at">--env</span> HF_TOKEN</span>
-<span id="cb15-3"><a href="#cb15-3" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb15-4"><a href="#cb15-4" aria-hidden="true" tabindex="-1"></a><span class="co"># Managed spot (auto-recovery on preemption)</span></span>
-<span id="cb15-5"><a href="#cb15-5" aria-hidden="true" tabindex="-1"></a><span class="va">HF_TOKEN</span><span class="op">=</span>xx <span class="va">BUCKET</span><span class="op">=&lt;</span>unique-name<span class="op">&gt;</span> sky <span class="ex">spot</span> launch axolotl-spot.yaml <span class="at">--env</span> HF_TOKEN <span class="at">--env</span> BUCKET</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb14"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb14-1"><a href="#cb14-1" aria-hidden="true" tabindex="-1"></a><span class="co"># On-demand</span></span>
+<span id="cb14-2"><a href="#cb14-2" aria-hidden="true" tabindex="-1"></a><span class="va">HF_TOKEN</span><span class="op">=</span>xx <span class="ex">sky</span> launch axolotl.yaml <span class="at">--env</span> HF_TOKEN</span>
+<span id="cb14-3"><a href="#cb14-3" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb14-4"><a href="#cb14-4" aria-hidden="true" tabindex="-1"></a><span class="co"># Managed spot (auto-recovery on preemption)</span></span>
+<span id="cb14-5"><a href="#cb14-5" aria-hidden="true" tabindex="-1"></a><span class="va">HF_TOKEN</span><span class="op">=</span>xx <span class="va">BUCKET</span><span class="op">=&lt;</span>unique-name<span class="op">&gt;</span> sky <span class="ex">spot</span> launch axolotl-spot.yaml <span class="at">--env</span> HF_TOKEN <span class="at">--env</span> BUCKET</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </section>
 </section>
 <section id="dataset" class="level3">
@@ -687,124 +680,82 @@ cd skypilot/llm/axolotl</code></pre>
 <p>See <a href="examples">examples</a> for quick start. It is recommended to duplicate and modify to your needs. The most important options are:</p>
 <ul>
 <li><p>model</p>
-<div class="sourceCode" id="cb16"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb16-1"><a href="#cb16-1" aria-hidden="true" tabindex="-1"></a><span class="fu">base_model</span><span class="kw">:</span><span class="at"> ./llama-7b-hf</span><span class="co"> # local or huggingface repo</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb15"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb15-1"><a href="#cb15-1" aria-hidden="true" tabindex="-1"></a><span class="fu">base_model</span><span class="kw">:</span><span class="at"> ./llama-7b-hf</span><span class="co"> # local or huggingface repo</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>Note: The code will load the right architecture.</p></li>
 <li><p>dataset</p>
-<div class="sourceCode" id="cb17"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb17-1"><a href="#cb17-1" aria-hidden="true" tabindex="-1"></a><span class="fu">datasets</span><span class="kw">:</span></span>
-<span id="cb17-2"><a href="#cb17-2" aria-hidden="true" tabindex="-1"></a><span class="co">    # huggingface repo</span></span>
-<span id="cb17-3"><a href="#cb17-3" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> vicgalle/alpaca-gpt4</span></span>
-<span id="cb17-4"><a href="#cb17-4" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> alpaca</span></span>
-<span id="cb17-5"><a href="#cb17-5" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb17-6"><a href="#cb17-6" aria-hidden="true" tabindex="-1"></a><span class="co">    # huggingface repo with specific configuration/subset</span></span>
-<span id="cb17-7"><a href="#cb17-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> EleutherAI/pile</span></span>
-<span id="cb17-8"><a href="#cb17-8" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">name</span><span class="kw">:</span><span class="at"> enron_emails</span></span>
-<span id="cb17-9"><a href="#cb17-9" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> completion</span><span class="co"> # format from earlier</span></span>
-<span id="cb17-10"><a href="#cb17-10" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">field</span><span class="kw">:</span><span class="at"> text</span><span class="co"> # Optional[str] default: text, field to use for completion data</span></span>
-<span id="cb17-11"><a href="#cb17-11" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb17-12"><a href="#cb17-12" aria-hidden="true" tabindex="-1"></a><span class="co">    # huggingface repo with multiple named configurations/subsets</span></span>
-<span id="cb17-13"><a href="#cb17-13" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> bigcode/commitpackft</span></span>
-<span id="cb17-14"><a href="#cb17-14" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">name</span><span class="kw">:</span></span>
-<span id="cb17-15"><a href="#cb17-15" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="kw">-</span><span class="at"> ruby</span></span>
-<span id="cb17-16"><a href="#cb17-16" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="kw">-</span><span class="at"> python</span></span>
-<span id="cb17-17"><a href="#cb17-17" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="kw">-</span><span class="at"> typescript</span></span>
-<span id="cb17-18"><a href="#cb17-18" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> ...</span><span class="co"> # unimplemented custom format</span></span>
-<span id="cb17-19"><a href="#cb17-19" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb17-20"><a href="#cb17-20" aria-hidden="true" tabindex="-1"></a><span class="co">    # fastchat conversation</span></span>
-<span id="cb17-21"><a href="#cb17-21" aria-hidden="true" tabindex="-1"></a><span class="co">    # See 'conversation' options: https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py</span></span>
-<span id="cb17-22"><a href="#cb17-22" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> ...</span></span>
-<span id="cb17-23"><a href="#cb17-23" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> sharegpt</span></span>
-<span id="cb17-24"><a href="#cb17-24" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">conversation</span><span class="kw">:</span><span class="at"> chatml</span><span class="co"> # default: vicuna_v1.1</span></span>
-<span id="cb17-25"><a href="#cb17-25" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb17-26"><a href="#cb17-26" aria-hidden="true" tabindex="-1"></a><span class="co">    # local</span></span>
-<span id="cb17-27"><a href="#cb17-27" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> data.jsonl</span><span class="co"> # or json</span></span>
-<span id="cb17-28"><a href="#cb17-28" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">ds_type</span><span class="kw">:</span><span class="at"> json</span><span class="co"> # see other options below</span></span>
-<span id="cb17-29"><a href="#cb17-29" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> alpaca</span></span>
-<span id="cb17-30"><a href="#cb17-30" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb17-31"><a href="#cb17-31" aria-hidden="true" tabindex="-1"></a><span class="co">    # dataset with splits, but no train split</span></span>
-<span id="cb17-32"><a href="#cb17-32" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> knowrohit07/know_sql</span></span>
-<span id="cb17-33"><a href="#cb17-33" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> context_qa.load_v2</span></span>
-<span id="cb17-34"><a href="#cb17-34" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">train_on_split</span><span class="kw">:</span><span class="at"> validation</span></span>
-<span id="cb17-35"><a href="#cb17-35" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb17-36"><a href="#cb17-36" aria-hidden="true" tabindex="-1"></a><span class="co">    # loading from s3 or gcs</span></span>
-<span id="cb17-37"><a href="#cb17-37" aria-hidden="true" tabindex="-1"></a><span class="co">    # s3 creds will be loaded from the system default and gcs only supports public access</span></span>
-<span id="cb17-38"><a href="#cb17-38" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> s3://path_to_ds</span><span class="co"> # Accepts folder with arrow/parquet or file path like above. Supports s3, gcs.</span></span>
-<span id="cb17-39"><a href="#cb17-39" aria-hidden="true" tabindex="-1"></a><span class="at">    ...</span></span>
-<span id="cb17-40"><a href="#cb17-40" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb17-41"><a href="#cb17-41" aria-hidden="true" tabindex="-1"></a><span class="co">    # Loading Data From a Public URL</span></span>
-<span id="cb17-42"><a href="#cb17-42" aria-hidden="true" tabindex="-1"></a><span class="co">    # - The file format is `json` (which includes `jsonl`) by default. For different formats, adjust the `ds_type` option accordingly.</span></span>
-<span id="cb17-43"><a href="#cb17-43" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> https://some.url.com/yourdata.jsonl</span><span class="co"> # The URL should be a direct link to the file you wish to load. URLs must use HTTPS protocol, not HTTP.</span></span>
-<span id="cb17-44"><a href="#cb17-44" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">ds_type</span><span class="kw">:</span><span class="at"> json</span><span class="co"> # this is the default, see other options below.</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
+<div class="sourceCode" id="cb16"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb16-1"><a href="#cb16-1" aria-hidden="true" tabindex="-1"></a><span class="fu">datasets</span><span class="kw">:</span></span>
+<span id="cb16-2"><a href="#cb16-2" aria-hidden="true" tabindex="-1"></a><span class="co">    # huggingface repo</span></span>
+<span id="cb16-3"><a href="#cb16-3" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> vicgalle/alpaca-gpt4</span></span>
+<span id="cb16-4"><a href="#cb16-4" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> alpaca</span></span>
+<span id="cb16-5"><a href="#cb16-5" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb16-6"><a href="#cb16-6" aria-hidden="true" tabindex="-1"></a><span class="co">    # huggingface repo with specific configuration/subset</span></span>
+<span id="cb16-7"><a href="#cb16-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> EleutherAI/pile</span></span>
+<span id="cb16-8"><a href="#cb16-8" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">name</span><span class="kw">:</span><span class="at"> enron_emails</span></span>
+<span id="cb16-9"><a href="#cb16-9" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> completion</span><span class="co"> # format from earlier</span></span>
+<span id="cb16-10"><a href="#cb16-10" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">field</span><span class="kw">:</span><span class="at"> text</span><span class="co"> # Optional[str] default: text, field to use for completion data</span></span>
+<span id="cb16-11"><a href="#cb16-11" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb16-12"><a href="#cb16-12" aria-hidden="true" tabindex="-1"></a><span class="co">    # huggingface repo with multiple named configurations/subsets</span></span>
+<span id="cb16-13"><a href="#cb16-13" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> bigcode/commitpackft</span></span>
+<span id="cb16-14"><a href="#cb16-14" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">name</span><span class="kw">:</span></span>
+<span id="cb16-15"><a href="#cb16-15" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="kw">-</span><span class="at"> ruby</span></span>
+<span id="cb16-16"><a href="#cb16-16" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="kw">-</span><span class="at"> python</span></span>
+<span id="cb16-17"><a href="#cb16-17" aria-hidden="true" tabindex="-1"></a><span class="at">      </span><span class="kw">-</span><span class="at"> typescript</span></span>
+<span id="cb16-18"><a href="#cb16-18" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> ...</span><span class="co"> # unimplemented custom format</span></span>
+<span id="cb16-19"><a href="#cb16-19" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb16-20"><a href="#cb16-20" aria-hidden="true" tabindex="-1"></a><span class="co">    # fastchat conversation</span></span>
+<span id="cb16-21"><a href="#cb16-21" aria-hidden="true" tabindex="-1"></a><span class="co">    # See 'conversation' options: https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py</span></span>
+<span id="cb16-22"><a href="#cb16-22" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> ...</span></span>
+<span id="cb16-23"><a href="#cb16-23" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> sharegpt</span></span>
+<span id="cb16-24"><a href="#cb16-24" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">conversation</span><span class="kw">:</span><span class="at"> chatml</span><span class="co"> # default: vicuna_v1.1</span></span>
+<span id="cb16-25"><a href="#cb16-25" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb16-26"><a href="#cb16-26" aria-hidden="true" tabindex="-1"></a><span class="co">    # local</span></span>
+<span id="cb16-27"><a href="#cb16-27" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> data.jsonl</span><span class="co"> # or json</span></span>
+<span id="cb16-28"><a href="#cb16-28" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">ds_type</span><span class="kw">:</span><span class="at"> json</span><span class="co"> # see other options below</span></span>
+<span id="cb16-29"><a href="#cb16-29" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> alpaca</span></span>
+<span id="cb16-30"><a href="#cb16-30" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb16-31"><a href="#cb16-31" aria-hidden="true" tabindex="-1"></a><span class="co">    # dataset with splits, but no train split</span></span>
+<span id="cb16-32"><a href="#cb16-32" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> knowrohit07/know_sql</span></span>
+<span id="cb16-33"><a href="#cb16-33" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">type</span><span class="kw">:</span><span class="at"> context_qa.load_v2</span></span>
+<span id="cb16-34"><a href="#cb16-34" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">train_on_split</span><span class="kw">:</span><span class="at"> validation</span></span>
+<span id="cb16-35"><a href="#cb16-35" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb16-36"><a href="#cb16-36" aria-hidden="true" tabindex="-1"></a><span class="co">    # loading from s3 or gcs</span></span>
+<span id="cb16-37"><a href="#cb16-37" aria-hidden="true" tabindex="-1"></a><span class="co">    # s3 creds will be loaded from the system default and gcs only supports public access</span></span>
+<span id="cb16-38"><a href="#cb16-38" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> s3://path_to_ds</span><span class="co"> # Accepts folder with arrow/parquet or file path like above. Supports s3, gcs.</span></span>
+<span id="cb16-39"><a href="#cb16-39" aria-hidden="true" tabindex="-1"></a><span class="at">    ...</span></span>
+<span id="cb16-40"><a href="#cb16-40" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb16-41"><a href="#cb16-41" aria-hidden="true" tabindex="-1"></a><span class="co">    # Loading Data From a Public URL</span></span>
+<span id="cb16-42"><a href="#cb16-42" aria-hidden="true" tabindex="-1"></a><span class="co">    # - The file format is `json` (which includes `jsonl`) by default. For different formats, adjust the `ds_type` option accordingly.</span></span>
+<span id="cb16-43"><a href="#cb16-43" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="fu">path</span><span class="kw">:</span><span class="at"> https://some.url.com/yourdata.jsonl</span><span class="co"> # The URL should be a direct link to the file you wish to load. URLs must use HTTPS protocol, not HTTP.</span></span>
+<span id="cb16-44"><a href="#cb16-44" aria-hidden="true" tabindex="-1"></a><span class="at">    </span><span class="fu">ds_type</span><span class="kw">:</span><span class="at"> json</span><span class="co"> # this is the default, see other options below.</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
 <li><p>loading</p>
-<div class="sourceCode" id="cb18"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb18-1"><a href="#cb18-1" aria-hidden="true" tabindex="-1"></a><span class="fu">load_in_4bit</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span></span>
-<span id="cb18-2"><a href="#cb18-2" aria-hidden="true" tabindex="-1"></a><span class="fu">load_in_8bit</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span></span>
-<span id="cb18-3"><a href="#cb18-3" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb18-4"><a href="#cb18-4" aria-hidden="true" tabindex="-1"></a><span class="fu">bf16</span><span class="kw">:</span><span class="at"> auto</span><span class="co"> # require &gt;=ampere, auto will detect if your GPU supports this and choose automatically.</span></span>
-<span id="cb18-5"><a href="#cb18-5" aria-hidden="true" tabindex="-1"></a><span class="fu">fp16</span><span class="kw">:</span><span class="co"> # leave empty to use fp16 when bf16 is 'auto'. set to false if you want to fallback to fp32</span></span>
-<span id="cb18-6"><a href="#cb18-6" aria-hidden="true" tabindex="-1"></a><span class="fu">tf32</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span><span class="co"> # require &gt;=ampere</span></span>
-<span id="cb18-7"><a href="#cb18-7" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb18-8"><a href="#cb18-8" aria-hidden="true" tabindex="-1"></a><span class="fu">bfloat16</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span><span class="co"> # require &gt;=ampere, use instead of bf16 when you don't want AMP (automatic mixed precision)</span></span>
-<span id="cb18-9"><a href="#cb18-9" aria-hidden="true" tabindex="-1"></a><span class="fu">float16</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span><span class="co"> # use instead of fp16 when you don't want AMP</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb17"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb17-1"><a href="#cb17-1" aria-hidden="true" tabindex="-1"></a><span class="fu">load_in_4bit</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span></span>
+<span id="cb17-2"><a href="#cb17-2" aria-hidden="true" tabindex="-1"></a><span class="fu">load_in_8bit</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span></span>
+<span id="cb17-3"><a href="#cb17-3" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb17-4"><a href="#cb17-4" aria-hidden="true" tabindex="-1"></a><span class="fu">bf16</span><span class="kw">:</span><span class="at"> auto</span><span class="co"> # require &gt;=ampere, auto will detect if your GPU supports this and choose automatically.</span></span>
+<span id="cb17-5"><a href="#cb17-5" aria-hidden="true" tabindex="-1"></a><span class="fu">fp16</span><span class="kw">:</span><span class="co"> # leave empty to use fp16 when bf16 is 'auto'. set to false if you want to fallback to fp32</span></span>
+<span id="cb17-6"><a href="#cb17-6" aria-hidden="true" tabindex="-1"></a><span class="fu">tf32</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span><span class="co"> # require &gt;=ampere</span></span>
+<span id="cb17-7"><a href="#cb17-7" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb17-8"><a href="#cb17-8" aria-hidden="true" tabindex="-1"></a><span class="fu">bfloat16</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span><span class="co"> # require &gt;=ampere, use instead of bf16 when you don't want AMP (automatic mixed precision)</span></span>
+<span id="cb17-9"><a href="#cb17-9" aria-hidden="true" tabindex="-1"></a><span class="fu">float16</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span><span class="co"> # use instead of fp16 when you don't want AMP</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>Note: Repo does not do 4-bit quantization.</p></li>
 <li><p>lora</p>
-<div class="sourceCode" id="cb19"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb19-1"><a href="#cb19-1" aria-hidden="true" tabindex="-1"></a><span class="fu">adapter</span><span class="kw">:</span><span class="at"> lora</span><span class="co"> # 'qlora' or leave blank for full finetune</span></span>
-<span id="cb19-2"><a href="#cb19-2" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_r</span><span class="kw">:</span><span class="at"> </span><span class="dv">8</span></span>
-<span id="cb19-3"><a href="#cb19-3" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_alpha</span><span class="kw">:</span><span class="at"> </span><span class="dv">16</span></span>
-<span id="cb19-4"><a href="#cb19-4" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_dropout</span><span class="kw">:</span><span class="at"> </span><span class="fl">0.05</span></span>
-<span id="cb19-5"><a href="#cb19-5" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_target_modules</span><span class="kw">:</span></span>
-<span id="cb19-6"><a href="#cb19-6" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> q_proj</span></span>
-<span id="cb19-7"><a href="#cb19-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> v_proj</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
+<div class="sourceCode" id="cb18"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb18-1"><a href="#cb18-1" aria-hidden="true" tabindex="-1"></a><span class="fu">adapter</span><span class="kw">:</span><span class="at"> lora</span><span class="co"> # 'qlora' or leave blank for full finetune</span></span>
+<span id="cb18-2"><a href="#cb18-2" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_r</span><span class="kw">:</span><span class="at"> </span><span class="dv">8</span></span>
+<span id="cb18-3"><a href="#cb18-3" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_alpha</span><span class="kw">:</span><span class="at"> </span><span class="dv">16</span></span>
+<span id="cb18-4"><a href="#cb18-4" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_dropout</span><span class="kw">:</span><span class="at"> </span><span class="fl">0.05</span></span>
+<span id="cb18-5"><a href="#cb18-5" aria-hidden="true" tabindex="-1"></a><span class="fu">lora_target_modules</span><span class="kw">:</span></span>
+<span id="cb18-6"><a href="#cb18-6" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> q_proj</span></span>
+<span id="cb18-7"><a href="#cb18-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> v_proj</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
 </ul>
 <section id="all-config-options" class="level4">
 <h4 class="anchored" data-anchor-id="all-config-options">All Config Options</h4>
 <p>See <a href="./docs/config.html">these docs</a> for all config options.</p>
-<details>
-<summary>
-Understanding of batch size and gradient accumulation steps
-</summary>
-<p><br> Gradient accumulation means accumulating gradients over several mini-batches and updating the model weights afterward. When the samples in each batch are diverse, this technique doesn’t significantly impact learning.</p>
-<p>This method allows for effective training with larger effective batch sizes without needing proportionally larger memory. Here’s why:</p>
-<ol type="1">
-<li><p><strong>Memory Consumption with Batch Size</strong>: The primary reason increasing the batch size impacts memory is due to the storage requirements for intermediate activations. When you forward propagate a batch through a network, you have to store the activations at each layer for each sample in the batch, because these activations are used during backpropagation to compute gradients. Therefore, larger batches mean more activations, leading to greater GPU memory consumption.</p></li>
-<li><p><strong>Gradient Accumulation</strong>: With gradient accumulation, you’re effectively simulating a larger batch size by accumulating gradients over several smaller batches (or micro-batches). However, at any given time, you’re only forward and backward propagating a micro-batch. This means you only store activations for the micro-batch, not the full accumulated batch. As a result, you can simulate the effect of a larger batch size without the memory cost of storing activations for a large batch.</p></li>
-</ol>
-<p><strong>Example 1:</strong> Micro batch size: 3 Gradient accumulation steps: 2 Number of GPUs: 3 Total batch size = 3 * 2 * 3 = 18</p>
-<pre><code>| GPU 1          | GPU 2          | GPU 3          |
-|----------------|----------------|----------------|
-| S1, S2, S3     | S4, S5, S6     | S7, S8, S9     |
-| e1, e2, e3     | e4, e5, e6     | e7, e8, e9     |
-|----------------|----------------|----------------|
-| → (accumulate) | → (accumulate) | → (accumulate) |
-|----------------|----------------|----------------|
-| S10, S11, S12  | S13, S14, S15  | S16, S17, S18  |
-| e10, e11, e12  | e13, e14, e15  | e16, e17, e18  |
-|----------------|----------------|----------------|
-| → (apply)      | → (apply)      | → (apply)      |
-
-Accumulated gradient for the weight w1 after the second iteration (considering all GPUs):
-Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6 + e7 + e8 + e9 + e10 + e11 + e12 + e13 + e14 + e15 + e16 + e17 + e18
-
-Weight update for w1:
-w1_new = w1_old - learning rate x (Total gradient for w1 / 18)</code></pre>
-<p><strong>Example 2:</strong> Micro batch size: 2 Gradient accumulation steps: 1 Number of GPUs: 3 Total batch size = 2 * 1 * 3 = 6</p>
-<pre><code>| GPU 1     | GPU 2     | GPU 3     |
-|-----------|-----------|-----------|
-| S1, S2    | S3, S4    | S5, S6    |
-| e1, e2    | e3, e4    | e5, e6    |
-|-----------|-----------|-----------|
-| → (apply) | → (apply) | → (apply) |
-
-Accumulated gradient for the weight w1 (considering all GPUs):
-Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6
-
-Weight update for w1:
-w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
-</details>
 </section>
 </section>
 <section id="train" class="level3">
 <h3 class="anchored" data-anchor-id="train">Train</h3>
 <p>Run</p>
-<div class="sourceCode" id="cb22"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb22-1"><a href="#cb22-1" aria-hidden="true" tabindex="-1"></a><span class="ex">accelerate</span> launch <span class="at">-m</span> axolotl.cli.train your_config.yml</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb19"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb19-1"><a href="#cb19-1" aria-hidden="true" tabindex="-1"></a><span class="ex">accelerate</span> launch <span class="at">-m</span> axolotl.cli.train your_config.yml</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <blockquote class="blockquote">
 <p>[!TIP] You can also reference a config file that is hosted on a public URL, for example <code>accelerate launch -m axolotl.cli.train https://yourdomain.com/your_config.yml</code></p>
 </blockquote>
@@ -816,7 +767,7 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <li>(Optional): Set <code>push_dataset_to_hub: hf_user/repo</code> to push it to Huggingface.</li>
 <li>(Optional): Use <code>--debug</code> to see preprocessed examples.</li>
 </ul>
-<div class="sourceCode" id="cb23"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb23-1"><a href="#cb23-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.preprocess your_config.yml</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb20"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb20-1"><a href="#cb20-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.preprocess your_config.yml</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </section>
 <section id="multi-gpu" class="level4">
 <h4 class="anchored" data-anchor-id="multi-gpu">Multi-GPU</h4>
@@ -825,7 +776,7 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <h5 class="anchored" data-anchor-id="deepspeed">DeepSpeed</h5>
 <p>Deepspeed is an optimization suite for multi-gpu systems allowing you to train much larger models than you might typically be able to fit into your GPU’s VRAM. More information about the various optimization types for deepspeed is available at https://huggingface.co/docs/accelerate/main/en/usage_guides/deepspeed#what-is-integrated</p>
 <p>We provide several default deepspeed JSON configurations for ZeRO stage 1, 2, and 3.</p>
-<div class="sourceCode" id="cb24"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb24-1"><a href="#cb24-1" aria-hidden="true" tabindex="-1"></a><span class="fu">deepspeed</span><span class="kw">:</span><span class="at"> deepspeed_configs/zero1.json</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb21"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb21-1"><a href="#cb21-1" aria-hidden="true" tabindex="-1"></a><span class="fu">deepspeed</span><span class="kw">:</span><span class="at"> deepspeed_configs/zero1.json</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <pre class="shell"><code>accelerate launch -m axolotl.cli.train examples/llama-2/config.py --deepspeed deepspeed_configs/zero1.json</code></pre>
 </section>
 <section id="fsdp" class="level5">
@@ -833,13 +784,13 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <ul>
 <li>llama FSDP</li>
 </ul>
-<div class="sourceCode" id="cb26"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb26-1"><a href="#cb26-1" aria-hidden="true" tabindex="-1"></a><span class="fu">fsdp</span><span class="kw">:</span></span>
-<span id="cb26-2"><a href="#cb26-2" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> full_shard</span></span>
-<span id="cb26-3"><a href="#cb26-3" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> auto_wrap</span></span>
-<span id="cb26-4"><a href="#cb26-4" aria-hidden="true" tabindex="-1"></a><span class="fu">fsdp_config</span><span class="kw">:</span></span>
-<span id="cb26-5"><a href="#cb26-5" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">fsdp_offload_params</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span></span>
-<span id="cb26-6"><a href="#cb26-6" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">fsdp_state_dict_type</span><span class="kw">:</span><span class="at"> FULL_STATE_DICT</span></span>
-<span id="cb26-7"><a href="#cb26-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">fsdp_transformer_layer_cls_to_wrap</span><span class="kw">:</span><span class="at"> LlamaDecoderLayer</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb23"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb23-1"><a href="#cb23-1" aria-hidden="true" tabindex="-1"></a><span class="fu">fsdp</span><span class="kw">:</span></span>
+<span id="cb23-2"><a href="#cb23-2" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> full_shard</span></span>
+<span id="cb23-3"><a href="#cb23-3" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> auto_wrap</span></span>
+<span id="cb23-4"><a href="#cb23-4" aria-hidden="true" tabindex="-1"></a><span class="fu">fsdp_config</span><span class="kw">:</span></span>
+<span id="cb23-5"><a href="#cb23-5" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">fsdp_offload_params</span><span class="kw">:</span><span class="at"> </span><span class="ch">true</span></span>
+<span id="cb23-6"><a href="#cb23-6" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">fsdp_state_dict_type</span><span class="kw">:</span><span class="at"> FULL_STATE_DICT</span></span>
+<span id="cb23-7"><a href="#cb23-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">fsdp_transformer_layer_cls_to_wrap</span><span class="kw">:</span><span class="at"> LlamaDecoderLayer</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </section>
 <section id="fsdp-qlora" class="level5">
 <h5 class="anchored" data-anchor-id="fsdp-qlora">FSDP + QLoRA</h5>
@@ -851,23 +802,23 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <ul>
 <li>wandb options</li>
 </ul>
-<div class="sourceCode" id="cb27"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb27-1"><a href="#cb27-1" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_mode</span><span class="kw">:</span></span>
-<span id="cb27-2"><a href="#cb27-2" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_project</span><span class="kw">:</span></span>
-<span id="cb27-3"><a href="#cb27-3" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_entity</span><span class="kw">:</span></span>
-<span id="cb27-4"><a href="#cb27-4" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_watch</span><span class="kw">:</span></span>
-<span id="cb27-5"><a href="#cb27-5" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_name</span><span class="kw">:</span></span>
-<span id="cb27-6"><a href="#cb27-6" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_log_model</span><span class="kw">:</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb24"><pre class="sourceCode yaml code-with-copy"><code class="sourceCode yaml"><span id="cb24-1"><a href="#cb24-1" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_mode</span><span class="kw">:</span></span>
+<span id="cb24-2"><a href="#cb24-2" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_project</span><span class="kw">:</span></span>
+<span id="cb24-3"><a href="#cb24-3" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_entity</span><span class="kw">:</span></span>
+<span id="cb24-4"><a href="#cb24-4" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_watch</span><span class="kw">:</span></span>
+<span id="cb24-5"><a href="#cb24-5" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_name</span><span class="kw">:</span></span>
+<span id="cb24-6"><a href="#cb24-6" aria-hidden="true" tabindex="-1"></a><span class="fu">wandb_log_model</span><span class="kw">:</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 </section>
 <section id="special-tokens" class="level5">
 <h5 class="anchored" data-anchor-id="special-tokens">Special Tokens</h5>
 <p>It is important to have special tokens like delimiters, end-of-sequence, beginning-of-sequence in your tokenizer’s vocabulary. This will help you avoid tokenization issues and help your model train better. You can do this in axolotl like this:</p>
-<div class="sourceCode" id="cb28"><pre class="sourceCode yml code-with-copy"><code class="sourceCode yaml"><span id="cb28-1"><a href="#cb28-1" aria-hidden="true" tabindex="-1"></a><span class="fu">special_tokens</span><span class="kw">:</span></span>
-<span id="cb28-2"><a href="#cb28-2" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">bos_token</span><span class="kw">:</span><span class="at"> </span><span class="st">"&lt;s&gt;"</span></span>
-<span id="cb28-3"><a href="#cb28-3" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">eos_token</span><span class="kw">:</span><span class="at"> </span><span class="st">"&lt;/s&gt;"</span></span>
-<span id="cb28-4"><a href="#cb28-4" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">unk_token</span><span class="kw">:</span><span class="at"> </span><span class="st">"&lt;unk&gt;"</span></span>
-<span id="cb28-5"><a href="#cb28-5" aria-hidden="true" tabindex="-1"></a><span class="fu">tokens</span><span class="kw">:</span><span class="co"> # these are delimiters</span></span>
-<span id="cb28-6"><a href="#cb28-6" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="st">"&lt;|im_start|&gt;"</span></span>
-<span id="cb28-7"><a href="#cb28-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="st">"&lt;|im_end|&gt;"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb25"><pre class="sourceCode yml code-with-copy"><code class="sourceCode yaml"><span id="cb25-1"><a href="#cb25-1" aria-hidden="true" tabindex="-1"></a><span class="fu">special_tokens</span><span class="kw">:</span></span>
+<span id="cb25-2"><a href="#cb25-2" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">bos_token</span><span class="kw">:</span><span class="at"> </span><span class="st">"&lt;s&gt;"</span></span>
+<span id="cb25-3"><a href="#cb25-3" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">eos_token</span><span class="kw">:</span><span class="at"> </span><span class="st">"&lt;/s&gt;"</span></span>
+<span id="cb25-4"><a href="#cb25-4" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="fu">unk_token</span><span class="kw">:</span><span class="at"> </span><span class="st">"&lt;unk&gt;"</span></span>
+<span id="cb25-5"><a href="#cb25-5" aria-hidden="true" tabindex="-1"></a><span class="fu">tokens</span><span class="kw">:</span><span class="co"> # these are delimiters</span></span>
+<span id="cb25-6"><a href="#cb25-6" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="st">"&lt;|im_start|&gt;"</span></span>
+<span id="cb25-7"><a href="#cb25-7" aria-hidden="true" tabindex="-1"></a><span class="at">  </span><span class="kw">-</span><span class="at"> </span><span class="st">"&lt;|im_end|&gt;"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>When you include these tokens in your axolotl config, axolotl adds these tokens to the tokenizer’s vocabulary.</p>
 </section>
 </section>
@@ -878,14 +829,14 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <p>Pass the appropriate flag to the inference command, depending upon what kind of model was trained:</p>
 <ul>
 <li><p>Pretrained LORA:</p>
-<div class="sourceCode" id="cb29"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb29-1"><a href="#cb29-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="at">--lora_model_dir</span><span class="op">=</span><span class="st">"./lora-output-dir"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
+<div class="sourceCode" id="cb26"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb26-1"><a href="#cb26-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="at">--lora_model_dir</span><span class="op">=</span><span class="st">"./lora-output-dir"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
 <li><p>Full weights finetune:</p>
-<div class="sourceCode" id="cb30"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb30-1"><a href="#cb30-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="at">--base_model</span><span class="op">=</span><span class="st">"./completed-model"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
+<div class="sourceCode" id="cb27"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb27-1"><a href="#cb27-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="at">--base_model</span><span class="op">=</span><span class="st">"./completed-model"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
 <li><p>Full weights finetune w/ a prompt from a text file:</p>
-<div class="sourceCode" id="cb31"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb31-1"><a href="#cb31-1" aria-hidden="true" tabindex="-1"></a><span class="fu">cat</span> /tmp/prompt.txt <span class="kw">|</span> <span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="dt">\</span></span>
-<span id="cb31-2"><a href="#cb31-2" aria-hidden="true" tabindex="-1"></a>  <span class="at">--base_model</span><span class="op">=</span><span class="st">"./completed-model"</span> <span class="at">--prompter</span><span class="op">=</span>None <span class="at">--load_in_8bit</span><span class="op">=</span>True</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb28"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb28-1"><a href="#cb28-1" aria-hidden="true" tabindex="-1"></a><span class="fu">cat</span> /tmp/prompt.txt <span class="kw">|</span> <span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="dt">\</span></span>
+<span id="cb28-2"><a href="#cb28-2" aria-hidden="true" tabindex="-1"></a>  <span class="at">--base_model</span><span class="op">=</span><span class="st">"./completed-model"</span> <span class="at">--prompter</span><span class="op">=</span>None <span class="at">--load_in_8bit</span><span class="op">=</span>True</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>– With gradio hosting</p>
-<div class="sourceCode" id="cb32"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb32-1"><a href="#cb32-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="at">--gradio</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
+<div class="sourceCode" id="cb29"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb29-1"><a href="#cb29-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python</span> <span class="at">-m</span> axolotl.cli.inference examples/your_config.yml <span class="at">--gradio</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div></li>
 </ul>
 <p>Please use <code>--sample_packing False</code> if you have it on and receive the error similar to below:</p>
 <blockquote class="blockquote">
@@ -895,9 +846,9 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <section id="merge-lora-to-base" class="level3">
 <h3 class="anchored" data-anchor-id="merge-lora-to-base">Merge LORA to base</h3>
 <p>The following command will merge your LORA adapater with your base model. You can optionally pass the argument <code>--lora_model_dir</code> to specify the directory where your LORA adapter was saved, otherwhise, this will be inferred from <code>output_dir</code> in your axolotl config file. The merged model is saved in the sub-directory <code>{lora_model_dir}/merged</code>.</p>
-<div class="sourceCode" id="cb33"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb33-1"><a href="#cb33-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python3</span> <span class="at">-m</span> axolotl.cli.merge_lora your_config.yml <span class="at">--lora_model_dir</span><span class="op">=</span><span class="st">"./completed-model"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb30"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb30-1"><a href="#cb30-1" aria-hidden="true" tabindex="-1"></a><span class="ex">python3</span> <span class="at">-m</span> axolotl.cli.merge_lora your_config.yml <span class="at">--lora_model_dir</span><span class="op">=</span><span class="st">"./completed-model"</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>You may need to use the <code>gpu_memory_limit</code> and/or <code>lora_on_cpu</code> config options to avoid running out of memory. If you still run out of CUDA memory, you can try to merge in system RAM with</p>
-<div class="sourceCode" id="cb34"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb34-1"><a href="#cb34-1" aria-hidden="true" tabindex="-1"></a><span class="va">CUDA_VISIBLE_DEVICES</span><span class="op">=</span><span class="st">""</span> <span class="ex">python3</span> <span class="at">-m</span> axolotl.cli.merge_lora ...</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb31"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb31-1"><a href="#cb31-1" aria-hidden="true" tabindex="-1"></a><span class="va">CUDA_VISIBLE_DEVICES</span><span class="op">=</span><span class="st">""</span> <span class="ex">python3</span> <span class="at">-m</span> axolotl.cli.merge_lora ...</span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>although this will be very slow, and using the config options above are recommended instead.</p>
 </section>
 </section>
@@ -955,7 +906,7 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <section id="badge" class="level2">
 <h2 class="anchored" data-anchor-id="badge">Badge ❤🏷️</h2>
 <p>Building something cool with Axolotl? Consider adding a badge to your model card.</p>
-<div class="sourceCode" id="cb35"><pre class="sourceCode markdown code-with-copy"><code class="sourceCode markdown"><span id="cb35-1"><a href="#cb35-1" aria-hidden="true" tabindex="-1"></a><span class="co">[</span><span class="ot">&lt;img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/&gt;</span><span class="co">](https://github.com/OpenAccess-AI-Collective/axolotl)</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<div class="sourceCode" id="cb32"><pre class="sourceCode markdown code-with-copy"><code class="sourceCode markdown"><span id="cb32-1"><a href="#cb32-1" aria-hidden="true" tabindex="-1"></a><span class="co">[</span><span class="ot">&lt;img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/&gt;</span><span class="co">](https://github.com/OpenAccess-AI-Collective/axolotl)</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p><a href="https://github.com/OpenAccess-AI-Collective/axolotl"><img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"></a></p>
 </section>
 <section id="community-showcase" class="level2">
@@ -969,21 +920,15 @@ w1_new = w1_old - learning rate × (Total gradient for w1 / 6)</code></pre>
 <p>Please read the <a href="./.github/CONTRIBUTING.md">contributing guide</a></p>
 <p>Bugs? Please check the <a href="https://github.com/OpenAccess-AI-Collective/axolotl/issues/bug">open issues</a> else create a new Issue.</p>
 <p>PRs are <strong>greatly welcome</strong>!</p>
-<p>Please run below to setup env</p>
-<div class="sourceCode" id="cb36"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb36-1"><a href="#cb36-1" aria-hidden="true" tabindex="-1"></a><span class="fu">git</span> clone https://github.com/OpenAccess-AI-Collective/axolotl</span>
-<span id="cb36-2"><a href="#cb36-2" aria-hidden="true" tabindex="-1"></a><span class="bu">cd</span> axolotl</span>
-<span id="cb36-3"><a href="#cb36-3" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-4"><a href="#cb36-4" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install packaging ninja</span>
-<span id="cb36-5"><a href="#cb36-5" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install <span class="at">-e</span> <span class="st">'.[flash-attn,deepspeed]'</span></span>
-<span id="cb36-6"><a href="#cb36-6" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-7"><a href="#cb36-7" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install <span class="at">-r</span> requirements-dev.txt <span class="at">-r</span> requirements-tests.txt</span>
-<span id="cb36-8"><a href="#cb36-8" aria-hidden="true" tabindex="-1"></a><span class="ex">pre-commit</span> install</span>
-<span id="cb36-9"><a href="#cb36-9" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-10"><a href="#cb36-10" aria-hidden="true" tabindex="-1"></a><span class="co"># test</span></span>
-<span id="cb36-11"><a href="#cb36-11" aria-hidden="true" tabindex="-1"></a><span class="ex">pytest</span> tests/</span>
-<span id="cb36-12"><a href="#cb36-12" aria-hidden="true" tabindex="-1"></a></span>
-<span id="cb36-13"><a href="#cb36-13" aria-hidden="true" tabindex="-1"></a><span class="co"># optional: run against all files</span></span>
-<span id="cb36-14"><a href="#cb36-14" aria-hidden="true" tabindex="-1"></a><span class="ex">pre-commit</span> run <span class="at">--all-files</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
+<p>Please run the quickstart instructions followed by the below to setup env:</p>
+<div class="sourceCode" id="cb33"><pre class="sourceCode bash code-with-copy"><code class="sourceCode bash"><span id="cb33-1"><a href="#cb33-1" aria-hidden="true" tabindex="-1"></a><span class="ex">pip3</span> install <span class="at">-r</span> requirements-dev.txt <span class="at">-r</span> requirements-tests.txt</span>
+<span id="cb33-2"><a href="#cb33-2" aria-hidden="true" tabindex="-1"></a><span class="ex">pre-commit</span> install</span>
+<span id="cb33-3"><a href="#cb33-3" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb33-4"><a href="#cb33-4" aria-hidden="true" tabindex="-1"></a><span class="co"># test</span></span>
+<span id="cb33-5"><a href="#cb33-5" aria-hidden="true" tabindex="-1"></a><span class="ex">pytest</span> tests/</span>
+<span id="cb33-6"><a href="#cb33-6" aria-hidden="true" tabindex="-1"></a></span>
+<span id="cb33-7"><a href="#cb33-7" aria-hidden="true" tabindex="-1"></a><span class="co"># optional: run against all files</span></span>
+<span id="cb33-8"><a href="#cb33-8" aria-hidden="true" tabindex="-1"></a><span class="ex">pre-commit</span> run <span class="at">--all-files</span></span></code><button title="Copy to Clipboard" class="code-copy-button"><i class="bi"></i></button></pre></div>
 <p>Thanks to all of our contributors to date. Help drive open source AI progress forward by contributing to Axolotl.</p>
 <p><a href="https://github.com/openaccess-ai-collective/axolotl/graphs/contributors"> <img src="https://contrib.rocks/image?repo=openaccess-ai-collective/axolotl" alt="contributor chart by https://contrib.rocks"> </a></p>
 </section>
diff --git a/listings.json b/listings.json
index 221f761d6..00fbd1fdb 100644
--- a/listings.json
+++ b/listings.json
@@ -2,9 +2,9 @@
   {
     "listing": "/docs/dataset-formats/index.html",
     "items": [
-      "/docs/dataset-formats/conversation.html",
-      "/docs/dataset-formats/inst_tune.html",
       "/docs/dataset-formats/pretraining.html",
+      "/docs/dataset-formats/inst_tune.html",
+      "/docs/dataset-formats/conversation.html",
       "/docs/dataset-formats/template_free.html",
       "/docs/dataset-formats/tokenized.html"
     ]
diff --git a/search.json b/search.json
index 2531f7b02..9dcb9c93a 100644
--- a/search.json
+++ b/search.json
@@ -64,7 +64,7 @@
     "href": "docs/input_output.html#background",
     "title": "Template-free prompt construction",
     "section": "Background",
-    "text": "Background\n\n\nMasking Inputs\nOne of the most popular features of axolotl is setting the following configuration value:\ntrain_on_inputs: false\nIf you declare a dataset formats such as alpaca or chatml, axolotl knows what is an input (i.e. human) vs. an output (i.e. the assistant) and masks the input labels so that your model can focus on predicting the outputs only.\n\n\n\nYou may not want prompt templates\nHowever, there are many situations where you don’t want to use one of these formats or templates (I usually don’t!). This is because they can:\n\nAdd unnecessary boilerplate to your prompts.\nCreate artifacts like special delimiters &lt;|im_start|&gt; that can quickly become footguns if you don’t include them correctly at inference time.\nEnforce a chat interface when you do not want one. Sometimes you just want to fine-tune a model to a very specific task and do NOT want multi-turn conversations, roles, etc.\nLimit you to only certain roles that the template allows.\n\n\n\n\nThe input_output format\nYou can construct your prompts without a template by using the input_output format, by setting type: input_output in your configuration file like this:\nconfig.yml\ntrain_on_inputs: false # Mask segments of your data\ndatasets:\n  - path: output.jsonl\n    type: input_output  # use template free prompt construction\nUnlike type: completion, which is also template-free, type: input_output allows you to mask segments of your text. More details on how this works are described below.",
+    "text": "Background\n\n\nMasking Inputs\nOne of the most popular features of axolotl is setting the following configuration value:\ntrain_on_inputs: false\nIf you declare a dataset formats such as alpaca or chatml, axolotl knows what is an input (i.e. human) vs. an output (i.e. the assistant) and masks the input labels so that your model can focus on predicting the outputs only.\n\n\n\nYou may not want prompt templates\nHowever, there are many situations where you don’t want to use one of these formats or templates. This is because they can:\n\nAdd unnecessary boilerplate to your prompts.\nCreate artifacts like special delimiters &lt;|im_start|&gt; that can quickly become footguns if you don’t include them correctly at inference time.\nEnforce a chat interface when you do not want one. Sometimes you just want to fine-tune a model to a very specific task and do NOT want multi-turn conversations, roles, etc.\nLimit you to only certain roles that the template allows.\n\n\n\n\nThe input_output format\nYou can construct your prompts without a template by using the input_output format, by setting type: input_output in your configuration file like this:\nconfig.yml\ntrain_on_inputs: false # Mask segments of your data\ndatasets:\n  - path: output.jsonl\n    type: input_output  # use template free prompt construction\nUnlike type: completion, which is also template-free, type: input_output allows you to mask segments of your text. More details on how this works are described below.",
     "crumbs": [
       "How-To Guides",
       "Template-free prompt construction"
@@ -400,66 +400,89 @@
       "Instruction Tuning"
     ]
   },
+  {
+    "objectID": "docs/dataset-formats/inst_tune.html#how-to-add-custom-prompt-format",
+    "href": "docs/dataset-formats/inst_tune.html#how-to-add-custom-prompt-format",
+    "title": "Instruction Tuning",
+    "section": "How to add custom prompt format",
+    "text": "How to add custom prompt format\nFor a dataset that is preprocessed for instruction purposes:\n\n\ndata.jsonl\n\n{\"input\": \"...\", \"output\": \"...\"}\n\nYou can use this example in your YAML config:\n\n\nconfig.yaml\n\ndatasets:\n  - path: repo\n    type:\n      system_prompt: \"\"\n      field_system: system\n      field_instruction: input\n      field_output: output\n      format: \"[INST] {instruction} [/INST]\"\n      no_input_format: \"[INST] {instruction} [/INST]\"\n\nSee full config options under here.",
+    "crumbs": [
+      "Dataset Formats",
+      "Instruction Tuning"
+    ]
+  },
   {
     "objectID": "docs/dataset-formats/conversation.html",
     "href": "docs/dataset-formats/conversation.html",
     "title": "Conversation",
     "section": "",
-    "text": "conversations where from is human/gpt. (optional: first row with role system to override default system prompt)\n\n\ndata.jsonl\n\n{\"conversations\": [{\"from\": \"...\", \"value\": \"...\"}]}\n\nNote: type: sharegpt opens a special config conversation: that enables conversions to many Conversation types. See the docs for all config options.\n\n\n\n\n\ndata.jsonl\n\n{\"conversations\": [{\"role\": \"...\", \"value\": \"...\"}]}\n\n\n\n\nconversations where role is used instead of from\n\n\ndata.jsonl\n\n{\"conversations\": [{\"role\": \"...\", \"value\": \"...\"}]}\n\n\n\n\nconversations where from is prompter assistant instead of default sharegpt\n\n\ndata.jsonl\n\n{\"conversations\": [{\"from\": \"...\", \"value\": \"...\"}]}\n\n\n\n\ncreates a chat where bot is asked to tell a joke, then explain why the joke is funny\n\n\ndata.jsonl\n\n{\"conversations\": [{\"title\": \"...\", \"text\": \"...\", \"explanation\": \"...\"}]}",
+    "text": "conversations where from is human/gpt. (optional: first row with role system to override default system prompt)\n\n\ndata.jsonl\n\n{\"conversations\": [{\"from\": \"...\", \"value\": \"...\"}]}\n\nNote: type: sharegpt opens special configs: - conversation: enables conversions to many Conversation types. Refer to the ‘name’ here for options. - roles: allows you to specify the roles for input and output. This is useful for datasets with custom roles such as tool etc to support masking. - field_human: specify the key to use instead of human in the conversation. - field_model: specify the key to use instead of gpt in the conversation.\ndatasets:\n    path: ...\n    type: sharegpt\n\n    conversation: # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py\n    field_human: # Optional[str]. Human key to use for conversation.\n    field_model: # Optional[str]. Assistant key to use for conversation.\n    # Add additional keys from your dataset as input or output roles\n    roles:\n      input: # Optional[List[str]]. These will be masked based on train_on_input\n      output: # Optional[List[str]].",
     "crumbs": [
       "Dataset Formats",
       "Conversation"
     ]
   },
   {
-    "objectID": "docs/dataset-formats/conversation.html#formats",
-    "href": "docs/dataset-formats/conversation.html#formats",
+    "objectID": "docs/dataset-formats/conversation.html#sharegpt",
+    "href": "docs/dataset-formats/conversation.html#sharegpt",
     "title": "Conversation",
     "section": "",
-    "text": "conversations where from is human/gpt. (optional: first row with role system to override default system prompt)\n\n\ndata.jsonl\n\n{\"conversations\": [{\"from\": \"...\", \"value\": \"...\"}]}\n\nNote: type: sharegpt opens a special config conversation: that enables conversions to many Conversation types. See the docs for all config options.\n\n\n\n\n\ndata.jsonl\n\n{\"conversations\": [{\"role\": \"...\", \"value\": \"...\"}]}\n\n\n\n\nconversations where role is used instead of from\n\n\ndata.jsonl\n\n{\"conversations\": [{\"role\": \"...\", \"value\": \"...\"}]}\n\n\n\n\nconversations where from is prompter assistant instead of default sharegpt\n\n\ndata.jsonl\n\n{\"conversations\": [{\"from\": \"...\", \"value\": \"...\"}]}\n\n\n\n\ncreates a chat where bot is asked to tell a joke, then explain why the joke is funny\n\n\ndata.jsonl\n\n{\"conversations\": [{\"title\": \"...\", \"text\": \"...\", \"explanation\": \"...\"}]}",
+    "text": "conversations where from is human/gpt. (optional: first row with role system to override default system prompt)\n\n\ndata.jsonl\n\n{\"conversations\": [{\"from\": \"...\", \"value\": \"...\"}]}\n\nNote: type: sharegpt opens special configs: - conversation: enables conversions to many Conversation types. Refer to the ‘name’ here for options. - roles: allows you to specify the roles for input and output. This is useful for datasets with custom roles such as tool etc to support masking. - field_human: specify the key to use instead of human in the conversation. - field_model: specify the key to use instead of gpt in the conversation.\ndatasets:\n    path: ...\n    type: sharegpt\n\n    conversation: # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py\n    field_human: # Optional[str]. Human key to use for conversation.\n    field_model: # Optional[str]. Assistant key to use for conversation.\n    # Add additional keys from your dataset as input or output roles\n    roles:\n      input: # Optional[List[str]]. These will be masked based on train_on_input\n      output: # Optional[List[str]].",
     "crumbs": [
       "Dataset Formats",
       "Conversation"
     ]
   },
   {
-    "objectID": "docs/dataset-formats/conversation.html#how-to-add-custom-prompts-for-instruction-tuning",
-    "href": "docs/dataset-formats/conversation.html#how-to-add-custom-prompts-for-instruction-tuning",
+    "objectID": "docs/dataset-formats/conversation.html#pygmalion",
+    "href": "docs/dataset-formats/conversation.html#pygmalion",
     "title": "Conversation",
-    "section": "How to add custom prompts for instruction-tuning",
-    "text": "How to add custom prompts for instruction-tuning\nFor a dataset that is preprocessed for instruction purposes:\n\n\ndata.jsonl\n\n{\"input\": \"...\", \"output\": \"...\"}\n\nYou can use this example in your YAML config:\n\n\nconfig.yaml\n\ndatasets:\n  - path: repo\n    type:\n      system_prompt: \"\"\n      field_system: system\n      field_instruction: input\n      field_output: output\n      format: \"[INST] {instruction} [/INST]\"\n      no_input_format: \"[INST] {instruction} [/INST]\"\n\nSee full config options under here.",
+    "section": "pygmalion",
+    "text": "pygmalion\n\n\ndata.jsonl\n\n{\"conversations\": [{\"role\": \"...\", \"value\": \"...\"}]}",
     "crumbs": [
       "Dataset Formats",
       "Conversation"
     ]
   },
   {
-    "objectID": "docs/rlhf.html",
-    "href": "docs/rlhf.html",
-    "title": "RLHF (Beta)",
-    "section": "",
-    "text": "Overview\nReinforcement Learning from Human Feedback is a method whereby a language model is optimized from data using human feedback. Various methods include, but not limited to:\n\nProximal Policy Optimization (PPO) (not yet supported in axolotl)\nDirect Preference Optimization (DPO)\nIdentity Preference Optimization (IPO)\n\n\n\nRLHF using Axolotl\n\n[!IMPORTANT] This is a BETA feature and many features are not fully implemented. You are encouraged to open new PRs to improve the integration and functionality.\n\nThe various RL training methods are implemented in trl and wrapped via axolotl. Below are various examples with how you can use various preference datasets to train models that use ChatML\n\nDPO\nrl: dpo\ndatasets:\n  - path: Intel/orca_dpo_pairs\n    split: train\n    type: chatml.intel\n  - path: argilla/ultrafeedback-binarized-preferences\n    split: train\n    type: chatml.argilla\n\n\nIPO\nrl: ipo\n\n\nORPO\nPaper: https://arxiv.org/abs/2403.07691\nrl: orpo\norpo_alpha: 0.1\nremove_unused_columns: false\n\nchat_template: chatml\ndatasets:\n  - path: argilla/ultrafeedback-binarized-preferences-cleaned\n    type: orpo.chat_template\n\n\nUsing local dataset files\ndatasets:\n  - ds_type: json\n    data_files:\n      - orca_rlhf.jsonl\n    split: train\n    type: chatml.intel\n\n\nTrl autounwrap for peft\nTrl supports autounwrapping peft models, so that a ref model does not need to be additionally loaded, leading to less VRAM needed. This is on by default. To turn it off, pass the following config.\n# load ref model when adapter training.\nrl_adapter_ref_model: true",
+    "objectID": "docs/dataset-formats/conversation.html#sharegpt.load_role",
+    "href": "docs/dataset-formats/conversation.html#sharegpt.load_role",
+    "title": "Conversation",
+    "section": "sharegpt.load_role",
+    "text": "sharegpt.load_role\nconversations where role is used instead of from\n\n\ndata.jsonl\n\n{\"conversations\": [{\"role\": \"...\", \"value\": \"...\"}]}",
     "crumbs": [
-      "How-To Guides",
-      "RLHF (Beta)"
+      "Dataset Formats",
+      "Conversation"
     ]
   },
   {
-    "objectID": "docs/faq.html",
-    "href": "docs/faq.html",
-    "title": "FAQ",
-    "section": "",
-    "text": "Q: The trainer stopped and hasn’t progressed in several minutes.\n\nA: Usually an issue with the GPUs communicating with each other. See the NCCL doc\n\nQ: Exitcode -9\n\nA: This usually happens when you run out of system RAM.\n\nQ: Exitcode -7 while using deepspeed\n\nA: Try upgrading deepspeed w: pip install -U deepspeed\n\nQ: AttributeError: ‘DummyOptim’ object has no attribute ‘step’\n\nA: You may be using deepspeed with single gpu. Please don’t set deepspeed: in yaml or cli.",
+    "objectID": "docs/dataset-formats/conversation.html#sharegpt.load_guanaco",
+    "href": "docs/dataset-formats/conversation.html#sharegpt.load_guanaco",
+    "title": "Conversation",
+    "section": "sharegpt.load_guanaco",
+    "text": "sharegpt.load_guanaco\nconversations where from is prompter assistant instead of default sharegpt\n\n\ndata.jsonl\n\n{\"conversations\": [{\"from\": \"...\", \"value\": \"...\"}]}",
     "crumbs": [
-      "FAQ"
+      "Dataset Formats",
+      "Conversation"
     ]
   },
   {
-    "objectID": "FAQS.html",
-    "href": "FAQS.html",
-    "title": "FAQs",
+    "objectID": "docs/dataset-formats/conversation.html#sharegpt_jokes",
+    "href": "docs/dataset-formats/conversation.html#sharegpt_jokes",
+    "title": "Conversation",
+    "section": "sharegpt_jokes",
+    "text": "sharegpt_jokes\ncreates a chat where bot is asked to tell a joke, then explain why the joke is funny\n\n\ndata.jsonl\n\n{\"conversations\": [{\"title\": \"...\", \"text\": \"...\", \"explanation\": \"...\"}]}",
+    "crumbs": [
+      "Dataset Formats",
+      "Conversation"
+    ]
+  },
+  {
+    "objectID": "docs/batch_vs_grad.html",
+    "href": "docs/batch_vs_grad.html",
+    "title": "Batch size vs Gradient accumulation",
     "section": "",
-    "text": "FAQs\n\nCan you train StableLM with this? Yes, but only with a single GPU atm. Multi GPU support is coming soon! Just waiting on this PR\nWill this work with Deepspeed? That’s still a WIP, but setting export ACCELERATE_USE_DEEPSPEED=true should work in some cases\nError invalid argument at line 359 in file /workspace/bitsandbytes/csrc/pythonInterface.c /arrow/cpp/src/arrow/filesystem/s3fs.cc:2598:  arrow::fs::FinalizeS3 was not called even though S3 was initialized. This could lead to a segmentation fault at exit. Try reinstalling bitsandbytes and transformers from source."
+    "text": "Gradient accumulation means accumulating gradients over several mini-batches and updating the model weights afterward. When the samples in each batch are diverse, this technique doesn’t significantly impact learning.\nThis method allows for effective training with larger effective batch sizes without needing proportionally larger memory. Here’s why:\n\nMemory Consumption with Batch Size: The primary reason increasing the batch size impacts memory is due to the storage requirements for intermediate activations. When you forward propagate a batch through a network, you have to store the activations at each layer for each sample in the batch, because these activations are used during backpropagation to compute gradients. Therefore, larger batches mean more activations, leading to greater GPU memory consumption.\nGradient Accumulation: With gradient accumulation, you’re effectively simulating a larger batch size by accumulating gradients over several smaller batches (or micro-batches). However, at any given time, you’re only forward and backward propagating a micro-batch. This means you only store activations for the micro-batch, not the full accumulated batch. As a result, you can simulate the effect of a larger batch size without the memory cost of storing activations for a large batch.\n\nExample 1: Micro batch size: 3 Gradient accumulation steps: 2 Number of GPUs: 3 Total batch size = 3 * 2 * 3 = 18\n| GPU 1          | GPU 2          | GPU 3          |\n|----------------|----------------|----------------|\n| S1, S2, S3     | S4, S5, S6     | S7, S8, S9     |\n| e1, e2, e3     | e4, e5, e6     | e7, e8, e9     |\n|----------------|----------------|----------------|\n| → (accumulate) | → (accumulate) | → (accumulate) |\n|----------------|----------------|----------------|\n| S10, S11, S12  | S13, S14, S15  | S16, S17, S18  |\n| e10, e11, e12  | e13, e14, e15  | e16, e17, e18  |\n|----------------|----------------|----------------|\n| → (apply)      | → (apply)      | → (apply)      |\n\nAccumulated gradient for the weight w1 after the second iteration (considering all GPUs):\nTotal gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6 + e7 + e8 + e9 + e10 + e11 + e12 + e13 + e14 + e15 + e16 + e17 + e18\n\nWeight update for w1:\nw1_new = w1_old - learning rate x (Total gradient for w1 / 18)\nExample 2: Micro batch size: 2 Gradient accumulation steps: 1 Number of GPUs: 3 Total batch size = 2 * 1 * 3 = 6\n| GPU 1     | GPU 2     | GPU 3     |\n|-----------|-----------|-----------|\n| S1, S2    | S3, S4    | S5, S6    |\n| e1, e2    | e3, e4    | e5, e6    |\n|-----------|-----------|-----------|\n| → (apply) | → (apply) | → (apply) |\n\nAccumulated gradient for the weight w1 (considering all GPUs):\nTotal gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6\n\nWeight update for w1:\nw1_new = w1_old - learning rate × (Total gradient for w1 / 6)"
   },
   {
     "objectID": "docs/fsdp_qlora.html",
@@ -527,6 +550,34 @@
       "FDSP + QLoRA"
     ]
   },
+  {
+    "objectID": "FAQS.html",
+    "href": "FAQS.html",
+    "title": "FAQs",
+    "section": "",
+    "text": "FAQs\n\nCan you train StableLM with this? Yes, but only with a single GPU atm. Multi GPU support is coming soon! Just waiting on this PR\nWill this work with Deepspeed? That’s still a WIP, but setting export ACCELERATE_USE_DEEPSPEED=true should work in some cases\nError invalid argument at line 359 in file /workspace/bitsandbytes/csrc/pythonInterface.c /arrow/cpp/src/arrow/filesystem/s3fs.cc:2598:  arrow::fs::FinalizeS3 was not called even though S3 was initialized. This could lead to a segmentation fault at exit. Try reinstalling bitsandbytes and transformers from source."
+  },
+  {
+    "objectID": "docs/faq.html",
+    "href": "docs/faq.html",
+    "title": "FAQ",
+    "section": "",
+    "text": "Q: The trainer stopped and hasn’t progressed in several minutes.\n\nA: Usually an issue with the GPUs communicating with each other. See the NCCL doc\n\nQ: Exitcode -9\n\nA: This usually happens when you run out of system RAM.\n\nQ: Exitcode -7 while using deepspeed\n\nA: Try upgrading deepspeed w: pip install -U deepspeed\n\nQ: AttributeError: ‘DummyOptim’ object has no attribute ‘step’\n\nA: You may be using deepspeed with single gpu. Please don’t set deepspeed: in yaml or cli.",
+    "crumbs": [
+      "FAQ"
+    ]
+  },
+  {
+    "objectID": "docs/rlhf.html",
+    "href": "docs/rlhf.html",
+    "title": "RLHF (Beta)",
+    "section": "",
+    "text": "Overview\nReinforcement Learning from Human Feedback is a method whereby a language model is optimized from data using human feedback. Various methods include, but not limited to:\n\nProximal Policy Optimization (PPO) (not yet supported in axolotl)\nDirect Preference Optimization (DPO)\nIdentity Preference Optimization (IPO)\n\n\n\nRLHF using Axolotl\n\n[!IMPORTANT] This is a BETA feature and many features are not fully implemented. You are encouraged to open new PRs to improve the integration and functionality.\n\nThe various RL training methods are implemented in trl and wrapped via axolotl. Below are various examples with how you can use various preference datasets to train models that use ChatML\n\nDPO\nrl: dpo\ndatasets:\n  - path: Intel/orca_dpo_pairs\n    split: train\n    type: chatml.intel\n  - path: argilla/ultrafeedback-binarized-preferences\n    split: train\n    type: chatml.argilla\n\n\nIPO\nrl: ipo\n\n\nORPO\nPaper: https://arxiv.org/abs/2403.07691\nrl: orpo\norpo_alpha: 0.1\nremove_unused_columns: false\n\nchat_template: chatml\ndatasets:\n  - path: argilla/ultrafeedback-binarized-preferences-cleaned\n    type: orpo.chat_template\n\n\nUsing local dataset files\ndatasets:\n  - ds_type: json\n    data_files:\n      - orca_rlhf.jsonl\n    split: train\n    type: chatml.intel\n\n\nTrl autounwrap for peft\nTrl supports autounwrapping peft models, so that a ref model does not need to be additionally loaded, leading to less VRAM needed. This is on by default. To turn it off, pass the following config.\n# load ref model when adapter training.\nrl_adapter_ref_model: true",
+    "crumbs": [
+      "How-To Guides",
+      "RLHF (Beta)"
+    ]
+  },
   {
     "objectID": "docs/dataset-formats/pretraining.html",
     "href": "docs/dataset-formats/pretraining.html",
@@ -554,7 +605,7 @@
     "href": "docs/dataset-formats/index.html",
     "title": "Dataset Formats",
     "section": "",
-    "text": "Axolotl supports a variety of dataset formats. It is recommended to use a JSONL format. The schema of the JSONL depends upon the task and the prompt template you wish to use. Instead of a JSONL, you can also use a HuggingFace dataset with columns for each JSONL field.\nBelow are these various formats organized by task:\n\n\n\n\n\n\n\n\n\nTitle\n\n\nDescription\n\n\n\n\n\n\nConversation\n\n\nConversation format for supervised fine-tuning.\n\n\n\n\nInstruction Tuning\n\n\nInstruction tuning formats for supervised fine-tuning.\n\n\n\n\nPre-training\n\n\nData format for a pre-training completion task.\n\n\n\n\nTemplate-Free\n\n\nConstruct prompts without a template.\n\n\n\n\nCustom Pre-Tokenized Dataset\n\n\nHow to use a custom pre-tokenized dataset.\n\n\n\n\n\nNo matching items",
+    "text": "Axolotl supports a variety of dataset formats. It is recommended to use a JSONL format. The schema of the JSONL depends upon the task and the prompt template you wish to use. Instead of a JSONL, you can also use a HuggingFace dataset with columns for each JSONL field.\nBelow are these various formats organized by task:\n\n\n\n\n\n\n\n\n\nTitle\n\n\nDescription\n\n\n\n\n\n\nPre-training\n\n\nData format for a pre-training completion task.\n\n\n\n\nInstruction Tuning\n\n\nInstruction tuning formats for supervised fine-tuning.\n\n\n\n\nConversation\n\n\nConversation format for supervised fine-tuning.\n\n\n\n\nTemplate-Free\n\n\nConstruct prompts without a template.\n\n\n\n\nCustom Pre-Tokenized Dataset\n\n\nHow to use a custom pre-tokenized dataset.\n\n\n\n\n\nNo matching items",
     "crumbs": [
       "Dataset Formats"
     ]
@@ -671,7 +722,7 @@
     "href": "index.html#advanced-setup",
     "title": "Axolotl",
     "section": "Advanced Setup",
-    "text": "Advanced Setup\n\nEnvironment\n\nDocker\ndocker run --gpus '\"all\"' --rm -it winglian/axolotl:main-latest\nOr run on the current files for development:\ndocker compose up -d\n\n[!Tip] If you want to debug axolotl or prefer to use Docker as your development environment, see the debugging guide’s section on Docker.\n\n\n\nDocker advanced\n\nA more powerful Docker command to run would be this:\ndocker run --privileged --gpus '\"all\"' --shm-size 10g --rm -it --name axolotl --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 --mount type=bind,src=\"${PWD}\",target=/workspace/axolotl -v ${HOME}/.cache/huggingface:/root/.cache/huggingface winglian/axolotl:main-latest\nIt additionally: * Prevents memory issues when running e.g. deepspeed (e.g. you could hit SIGBUS/signal 7 error) through --ipc and --ulimit args. * Persists the downloaded HF data (models etc.) and your modifications to axolotl code through --mount/-v args. * The --name argument simply makes it easier to refer to the container in vscode (Dev Containers: Attach to Running Container...) or in your terminal. * The --privileged flag gives all capabilities to the container. * The --shm-size 10g argument increases the shared memory size. Use this if you see exitcode: -7 errors using deepspeed.\nMore information on nvidia website\n\n\n\nConda/Pip venv\n\nInstall python &gt;=3.10\nInstall pytorch stable https://pytorch.org/get-started/locally/\nInstall Axolotl along with python dependencies bash         pip3 install packaging         pip3 install -e '.[flash-attn,deepspeed]'\n(Optional) Login to Huggingface to use gated models/datasets. bash         huggingface-cli login Get the token at huggingface.co/settings/tokens\n\n\n\nCloud GPU\nFor cloud GPU providers that support docker images, use winglian/axolotl-cloud:main-latest\n\non Latitude.sh use this direct link\non JarvisLabs.ai use this direct link\non RunPod use this direct link\n\n\n\nBare Metal Cloud GPU\n\nLambdaLabs\n\n\nClick to Expand\n\n\nInstall python\n\nsudo apt update\nsudo apt install -y python3.10\n\nsudo update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1\nsudo update-alternatives --config python # pick 3.10 if given option\npython -V # should be 3.10\n\nInstall pip\n\nwget https://bootstrap.pypa.io/get-pip.py\npython get-pip.py\n\nInstall torch\n\npip3 install -U torch --index-url https://download.pytorch.org/whl/cu118\n\nAxolotl\n\ngit clone https://github.com/OpenAccess-AI-Collective/axolotl\ncd axolotl\n\npip3 install packaging\npip3 install -e '.[flash-attn,deepspeed]'\npip3 install protobuf==3.20.3\npip3 install -U --ignore-installed requests Pillow psutil scipy\n\nSet path\n\nexport LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH\n\n\n\nGCP\n\n\nClick to Expand\n\nUse a Deeplearning linux OS with cuda and pytorch installed. Then follow instructions on quickstart.\nMake sure to run the below to uninstall xla.\npip uninstall -y torch_xla[tpu]\n\n\n\n\nWindows\nPlease use WSL or Docker!\n\n\nMac\nUse the below instead of the install method in QuickStart.\npip3 install -e '.'\nMore info: mac.md\n\n\nGoogle Colab\nPlease use this example notebook.\n\n\nLaunching on public clouds via SkyPilot\nTo launch on GPU instances (both on-demand and spot instances) on 7+ clouds (GCP, AWS, Azure, OCI, and more), you can use SkyPilot:\npip install \"skypilot-nightly[gcp,aws,azure,oci,lambda,kubernetes,ibm,scp]\"  # choose your clouds\nsky check\nGet the example YAMLs of using Axolotl to finetune mistralai/Mistral-7B-v0.1:\ngit clone https://github.com/skypilot-org/skypilot.git\ncd skypilot/llm/axolotl\nUse one command to launch:\n# On-demand\nHF_TOKEN=xx sky launch axolotl.yaml --env HF_TOKEN\n\n# Managed spot (auto-recovery on preemption)\nHF_TOKEN=xx BUCKET=&lt;unique-name&gt; sky spot launch axolotl-spot.yaml --env HF_TOKEN --env BUCKET\n\n\n\nDataset\nAxolotl supports a variety of dataset formats. It is recommended to use a JSONL. The schema of the JSONL depends upon the task and the prompt template you wish to use. Instead of a JSONL, you can also use a HuggingFace dataset with columns for each JSONL field.\nSee these docs for more information on how to use different dataset formats.\n\n\nConfig\nSee examples for quick start. It is recommended to duplicate and modify to your needs. The most important options are:\n\nmodel\nbase_model: ./llama-7b-hf # local or huggingface repo\nNote: The code will load the right architecture.\ndataset\ndatasets:\n    # huggingface repo\n  - path: vicgalle/alpaca-gpt4\n    type: alpaca\n\n    # huggingface repo with specific configuration/subset\n  - path: EleutherAI/pile\n    name: enron_emails\n    type: completion # format from earlier\n    field: text # Optional[str] default: text, field to use for completion data\n\n    # huggingface repo with multiple named configurations/subsets\n  - path: bigcode/commitpackft\n    name:\n      - ruby\n      - python\n      - typescript\n    type: ... # unimplemented custom format\n\n    # fastchat conversation\n    # See 'conversation' options: https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py\n  - path: ...\n    type: sharegpt\n    conversation: chatml # default: vicuna_v1.1\n\n    # local\n  - path: data.jsonl # or json\n    ds_type: json # see other options below\n    type: alpaca\n\n    # dataset with splits, but no train split\n  - path: knowrohit07/know_sql\n    type: context_qa.load_v2\n    train_on_split: validation\n\n    # loading from s3 or gcs\n    # s3 creds will be loaded from the system default and gcs only supports public access\n  - path: s3://path_to_ds # Accepts folder with arrow/parquet or file path like above. Supports s3, gcs.\n    ...\n\n    # Loading Data From a Public URL\n    # - The file format is `json` (which includes `jsonl`) by default. For different formats, adjust the `ds_type` option accordingly.\n  - path: https://some.url.com/yourdata.jsonl # The URL should be a direct link to the file you wish to load. URLs must use HTTPS protocol, not HTTP.\n    ds_type: json # this is the default, see other options below.\nloading\nload_in_4bit: true\nload_in_8bit: true\n\nbf16: auto # require &gt;=ampere, auto will detect if your GPU supports this and choose automatically.\nfp16: # leave empty to use fp16 when bf16 is 'auto'. set to false if you want to fallback to fp32\ntf32: true # require &gt;=ampere\n\nbfloat16: true # require &gt;=ampere, use instead of bf16 when you don't want AMP (automatic mixed precision)\nfloat16: true # use instead of fp16 when you don't want AMP\nNote: Repo does not do 4-bit quantization.\nlora\nadapter: lora # 'qlora' or leave blank for full finetune\nlora_r: 8\nlora_alpha: 16\nlora_dropout: 0.05\nlora_target_modules:\n  - q_proj\n  - v_proj\n\n\nAll Config Options\nSee these docs for all config options.\n\n\nUnderstanding of batch size and gradient accumulation steps\n\n Gradient accumulation means accumulating gradients over several mini-batches and updating the model weights afterward. When the samples in each batch are diverse, this technique doesn’t significantly impact learning.\nThis method allows for effective training with larger effective batch sizes without needing proportionally larger memory. Here’s why:\n\nMemory Consumption with Batch Size: The primary reason increasing the batch size impacts memory is due to the storage requirements for intermediate activations. When you forward propagate a batch through a network, you have to store the activations at each layer for each sample in the batch, because these activations are used during backpropagation to compute gradients. Therefore, larger batches mean more activations, leading to greater GPU memory consumption.\nGradient Accumulation: With gradient accumulation, you’re effectively simulating a larger batch size by accumulating gradients over several smaller batches (or micro-batches). However, at any given time, you’re only forward and backward propagating a micro-batch. This means you only store activations for the micro-batch, not the full accumulated batch. As a result, you can simulate the effect of a larger batch size without the memory cost of storing activations for a large batch.\n\nExample 1: Micro batch size: 3 Gradient accumulation steps: 2 Number of GPUs: 3 Total batch size = 3 * 2 * 3 = 18\n| GPU 1          | GPU 2          | GPU 3          |\n|----------------|----------------|----------------|\n| S1, S2, S3     | S4, S5, S6     | S7, S8, S9     |\n| e1, e2, e3     | e4, e5, e6     | e7, e8, e9     |\n|----------------|----------------|----------------|\n| → (accumulate) | → (accumulate) | → (accumulate) |\n|----------------|----------------|----------------|\n| S10, S11, S12  | S13, S14, S15  | S16, S17, S18  |\n| e10, e11, e12  | e13, e14, e15  | e16, e17, e18  |\n|----------------|----------------|----------------|\n| → (apply)      | → (apply)      | → (apply)      |\n\nAccumulated gradient for the weight w1 after the second iteration (considering all GPUs):\nTotal gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6 + e7 + e8 + e9 + e10 + e11 + e12 + e13 + e14 + e15 + e16 + e17 + e18\n\nWeight update for w1:\nw1_new = w1_old - learning rate x (Total gradient for w1 / 18)\nExample 2: Micro batch size: 2 Gradient accumulation steps: 1 Number of GPUs: 3 Total batch size = 2 * 1 * 3 = 6\n| GPU 1     | GPU 2     | GPU 3     |\n|-----------|-----------|-----------|\n| S1, S2    | S3, S4    | S5, S6    |\n| e1, e2    | e3, e4    | e5, e6    |\n|-----------|-----------|-----------|\n| → (apply) | → (apply) | → (apply) |\n\nAccumulated gradient for the weight w1 (considering all GPUs):\nTotal gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6\n\nWeight update for w1:\nw1_new = w1_old - learning rate × (Total gradient for w1 / 6)\n\n\n\n\nTrain\nRun\naccelerate launch -m axolotl.cli.train your_config.yml\n\n[!TIP] You can also reference a config file that is hosted on a public URL, for example accelerate launch -m axolotl.cli.train https://yourdomain.com/your_config.yml\n\n\nPreprocess dataset\nYou can optionally pre-tokenize dataset with the following before finetuning. This is recommended for large datasets.\n\nSet dataset_prepared_path: to a local folder for saving and loading pre-tokenized dataset.\n(Optional): Set push_dataset_to_hub: hf_user/repo to push it to Huggingface.\n(Optional): Use --debug to see preprocessed examples.\n\npython -m axolotl.cli.preprocess your_config.yml\n\n\nMulti-GPU\nBelow are the options available in axolotl for training with multiple GPUs. Note that DeepSpeed is the recommended multi-GPU option currently because FSDP may experience loss instability.\n\nDeepSpeed\nDeepspeed is an optimization suite for multi-gpu systems allowing you to train much larger models than you might typically be able to fit into your GPU’s VRAM. More information about the various optimization types for deepspeed is available at https://huggingface.co/docs/accelerate/main/en/usage_guides/deepspeed#what-is-integrated\nWe provide several default deepspeed JSON configurations for ZeRO stage 1, 2, and 3.\ndeepspeed: deepspeed_configs/zero1.json\naccelerate launch -m axolotl.cli.train examples/llama-2/config.py --deepspeed deepspeed_configs/zero1.json\n\n\nFSDP\n\nllama FSDP\n\nfsdp:\n  - full_shard\n  - auto_wrap\nfsdp_config:\n  fsdp_offload_params: true\n  fsdp_state_dict_type: FULL_STATE_DICT\n  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer\n\n\nFSDP + QLoRA\nAxolotl supports training with FSDP and QLoRA, see these docs for more information.\n\n\nWeights & Biases Logging\nMake sure your WANDB_API_KEY environment variable is set (recommended) or you login to wandb with wandb login.\n\nwandb options\n\nwandb_mode:\nwandb_project:\nwandb_entity:\nwandb_watch:\nwandb_name:\nwandb_log_model:\n\n\nSpecial Tokens\nIt is important to have special tokens like delimiters, end-of-sequence, beginning-of-sequence in your tokenizer’s vocabulary. This will help you avoid tokenization issues and help your model train better. You can do this in axolotl like this:\nspecial_tokens:\n  bos_token: \"&lt;s&gt;\"\n  eos_token: \"&lt;/s&gt;\"\n  unk_token: \"&lt;unk&gt;\"\ntokens: # these are delimiters\n  - \"&lt;|im_start|&gt;\"\n  - \"&lt;|im_end|&gt;\"\nWhen you include these tokens in your axolotl config, axolotl adds these tokens to the tokenizer’s vocabulary.\n\n\n\n\nInference Playground\nAxolotl allows you to load your model in an interactive terminal playground for quick experimentation. The config file is the same config file used for training.\nPass the appropriate flag to the inference command, depending upon what kind of model was trained:\n\nPretrained LORA:\npython -m axolotl.cli.inference examples/your_config.yml --lora_model_dir=\"./lora-output-dir\"\nFull weights finetune:\npython -m axolotl.cli.inference examples/your_config.yml --base_model=\"./completed-model\"\nFull weights finetune w/ a prompt from a text file:\ncat /tmp/prompt.txt | python -m axolotl.cli.inference examples/your_config.yml \\\n  --base_model=\"./completed-model\" --prompter=None --load_in_8bit=True\n– With gradio hosting\npython -m axolotl.cli.inference examples/your_config.yml --gradio\n\nPlease use --sample_packing False if you have it on and receive the error similar to below:\n\nRuntimeError: stack expects each tensor to be equal size, but got [1, 32, 1, 128] at entry 0 and [1, 32, 8, 128] at entry 1\n\n\n\nMerge LORA to base\nThe following command will merge your LORA adapater with your base model. You can optionally pass the argument --lora_model_dir to specify the directory where your LORA adapter was saved, otherwhise, this will be inferred from output_dir in your axolotl config file. The merged model is saved in the sub-directory {lora_model_dir}/merged.\npython3 -m axolotl.cli.merge_lora your_config.yml --lora_model_dir=\"./completed-model\"\nYou may need to use the gpu_memory_limit and/or lora_on_cpu config options to avoid running out of memory. If you still run out of CUDA memory, you can try to merge in system RAM with\nCUDA_VISIBLE_DEVICES=\"\" python3 -m axolotl.cli.merge_lora ...\nalthough this will be very slow, and using the config options above are recommended instead.",
+    "text": "Advanced Setup\n\nEnvironment\n\nDocker\ndocker run --gpus '\"all\"' --rm -it winglian/axolotl:main-latest\nOr run on the current files for development:\ndocker compose up -d\n\n[!Tip] If you want to debug axolotl or prefer to use Docker as your development environment, see the debugging guide’s section on Docker.\n\n\n\nDocker advanced\n\nA more powerful Docker command to run would be this:\ndocker run --privileged --gpus '\"all\"' --shm-size 10g --rm -it --name axolotl --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 --mount type=bind,src=\"${PWD}\",target=/workspace/axolotl -v ${HOME}/.cache/huggingface:/root/.cache/huggingface winglian/axolotl:main-latest\nIt additionally: * Prevents memory issues when running e.g. deepspeed (e.g. you could hit SIGBUS/signal 7 error) through --ipc and --ulimit args. * Persists the downloaded HF data (models etc.) and your modifications to axolotl code through --mount/-v args. * The --name argument simply makes it easier to refer to the container in vscode (Dev Containers: Attach to Running Container...) or in your terminal. * The --privileged flag gives all capabilities to the container. * The --shm-size 10g argument increases the shared memory size. Use this if you see exitcode: -7 errors using deepspeed.\nMore information on nvidia website\n\n\n\nConda/Pip venv\n\nInstall python &gt;=3.10\nInstall pytorch stable https://pytorch.org/get-started/locally/\nInstall Axolotl along with python dependencies bash         pip3 install packaging         pip3 install -e '.[flash-attn,deepspeed]'\n(Optional) Login to Huggingface to use gated models/datasets. bash         huggingface-cli login Get the token at huggingface.co/settings/tokens\n\n\n\nCloud GPU\nFor cloud GPU providers that support docker images, use winglian/axolotl-cloud:main-latest\n\non Latitude.sh use this direct link\non JarvisLabs.ai use this direct link\non RunPod use this direct link\n\n\n\nBare Metal Cloud GPU\n\nLambdaLabs\n\n\nClick to Expand\n\n\nInstall python\n\nsudo apt update\nsudo apt install -y python3.10\n\nsudo update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1\nsudo update-alternatives --config python # pick 3.10 if given option\npython -V # should be 3.10\n\nInstall pip\n\nwget https://bootstrap.pypa.io/get-pip.py\npython get-pip.py\n\nInstall Pytorch https://pytorch.org/get-started/locally/\nFollow instructions on quickstart.\nRun\n\npip3 install protobuf==3.20.3\npip3 install -U --ignore-installed requests Pillow psutil scipy\n\nSet path\n\nexport LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH\n\n\n\nGCP\n\n\nClick to Expand\n\nUse a Deeplearning linux OS with cuda and pytorch installed. Then follow instructions on quickstart.\nMake sure to run the below to uninstall xla.\npip uninstall -y torch_xla[tpu]\n\n\n\n\nWindows\nPlease use WSL or Docker!\n\n\nMac\nUse the below instead of the install method in QuickStart.\npip3 install -e '.'\nMore info: mac.md\n\n\nGoogle Colab\nPlease use this example notebook.\n\n\nLaunching on public clouds via SkyPilot\nTo launch on GPU instances (both on-demand and spot instances) on 7+ clouds (GCP, AWS, Azure, OCI, and more), you can use SkyPilot:\npip install \"skypilot-nightly[gcp,aws,azure,oci,lambda,kubernetes,ibm,scp]\"  # choose your clouds\nsky check\nGet the example YAMLs of using Axolotl to finetune mistralai/Mistral-7B-v0.1:\ngit clone https://github.com/skypilot-org/skypilot.git\ncd skypilot/llm/axolotl\nUse one command to launch:\n# On-demand\nHF_TOKEN=xx sky launch axolotl.yaml --env HF_TOKEN\n\n# Managed spot (auto-recovery on preemption)\nHF_TOKEN=xx BUCKET=&lt;unique-name&gt; sky spot launch axolotl-spot.yaml --env HF_TOKEN --env BUCKET\n\n\n\nDataset\nAxolotl supports a variety of dataset formats. It is recommended to use a JSONL. The schema of the JSONL depends upon the task and the prompt template you wish to use. Instead of a JSONL, you can also use a HuggingFace dataset with columns for each JSONL field.\nSee these docs for more information on how to use different dataset formats.\n\n\nConfig\nSee examples for quick start. It is recommended to duplicate and modify to your needs. The most important options are:\n\nmodel\nbase_model: ./llama-7b-hf # local or huggingface repo\nNote: The code will load the right architecture.\ndataset\ndatasets:\n    # huggingface repo\n  - path: vicgalle/alpaca-gpt4\n    type: alpaca\n\n    # huggingface repo with specific configuration/subset\n  - path: EleutherAI/pile\n    name: enron_emails\n    type: completion # format from earlier\n    field: text # Optional[str] default: text, field to use for completion data\n\n    # huggingface repo with multiple named configurations/subsets\n  - path: bigcode/commitpackft\n    name:\n      - ruby\n      - python\n      - typescript\n    type: ... # unimplemented custom format\n\n    # fastchat conversation\n    # See 'conversation' options: https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py\n  - path: ...\n    type: sharegpt\n    conversation: chatml # default: vicuna_v1.1\n\n    # local\n  - path: data.jsonl # or json\n    ds_type: json # see other options below\n    type: alpaca\n\n    # dataset with splits, but no train split\n  - path: knowrohit07/know_sql\n    type: context_qa.load_v2\n    train_on_split: validation\n\n    # loading from s3 or gcs\n    # s3 creds will be loaded from the system default and gcs only supports public access\n  - path: s3://path_to_ds # Accepts folder with arrow/parquet or file path like above. Supports s3, gcs.\n    ...\n\n    # Loading Data From a Public URL\n    # - The file format is `json` (which includes `jsonl`) by default. For different formats, adjust the `ds_type` option accordingly.\n  - path: https://some.url.com/yourdata.jsonl # The URL should be a direct link to the file you wish to load. URLs must use HTTPS protocol, not HTTP.\n    ds_type: json # this is the default, see other options below.\nloading\nload_in_4bit: true\nload_in_8bit: true\n\nbf16: auto # require &gt;=ampere, auto will detect if your GPU supports this and choose automatically.\nfp16: # leave empty to use fp16 when bf16 is 'auto'. set to false if you want to fallback to fp32\ntf32: true # require &gt;=ampere\n\nbfloat16: true # require &gt;=ampere, use instead of bf16 when you don't want AMP (automatic mixed precision)\nfloat16: true # use instead of fp16 when you don't want AMP\nNote: Repo does not do 4-bit quantization.\nlora\nadapter: lora # 'qlora' or leave blank for full finetune\nlora_r: 8\nlora_alpha: 16\nlora_dropout: 0.05\nlora_target_modules:\n  - q_proj\n  - v_proj\n\n\nAll Config Options\nSee these docs for all config options.\n\n\n\nTrain\nRun\naccelerate launch -m axolotl.cli.train your_config.yml\n\n[!TIP] You can also reference a config file that is hosted on a public URL, for example accelerate launch -m axolotl.cli.train https://yourdomain.com/your_config.yml\n\n\nPreprocess dataset\nYou can optionally pre-tokenize dataset with the following before finetuning. This is recommended for large datasets.\n\nSet dataset_prepared_path: to a local folder for saving and loading pre-tokenized dataset.\n(Optional): Set push_dataset_to_hub: hf_user/repo to push it to Huggingface.\n(Optional): Use --debug to see preprocessed examples.\n\npython -m axolotl.cli.preprocess your_config.yml\n\n\nMulti-GPU\nBelow are the options available in axolotl for training with multiple GPUs. Note that DeepSpeed is the recommended multi-GPU option currently because FSDP may experience loss instability.\n\nDeepSpeed\nDeepspeed is an optimization suite for multi-gpu systems allowing you to train much larger models than you might typically be able to fit into your GPU’s VRAM. More information about the various optimization types for deepspeed is available at https://huggingface.co/docs/accelerate/main/en/usage_guides/deepspeed#what-is-integrated\nWe provide several default deepspeed JSON configurations for ZeRO stage 1, 2, and 3.\ndeepspeed: deepspeed_configs/zero1.json\naccelerate launch -m axolotl.cli.train examples/llama-2/config.py --deepspeed deepspeed_configs/zero1.json\n\n\nFSDP\n\nllama FSDP\n\nfsdp:\n  - full_shard\n  - auto_wrap\nfsdp_config:\n  fsdp_offload_params: true\n  fsdp_state_dict_type: FULL_STATE_DICT\n  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer\n\n\nFSDP + QLoRA\nAxolotl supports training with FSDP and QLoRA, see these docs for more information.\n\n\nWeights & Biases Logging\nMake sure your WANDB_API_KEY environment variable is set (recommended) or you login to wandb with wandb login.\n\nwandb options\n\nwandb_mode:\nwandb_project:\nwandb_entity:\nwandb_watch:\nwandb_name:\nwandb_log_model:\n\n\nSpecial Tokens\nIt is important to have special tokens like delimiters, end-of-sequence, beginning-of-sequence in your tokenizer’s vocabulary. This will help you avoid tokenization issues and help your model train better. You can do this in axolotl like this:\nspecial_tokens:\n  bos_token: \"&lt;s&gt;\"\n  eos_token: \"&lt;/s&gt;\"\n  unk_token: \"&lt;unk&gt;\"\ntokens: # these are delimiters\n  - \"&lt;|im_start|&gt;\"\n  - \"&lt;|im_end|&gt;\"\nWhen you include these tokens in your axolotl config, axolotl adds these tokens to the tokenizer’s vocabulary.\n\n\n\n\nInference Playground\nAxolotl allows you to load your model in an interactive terminal playground for quick experimentation. The config file is the same config file used for training.\nPass the appropriate flag to the inference command, depending upon what kind of model was trained:\n\nPretrained LORA:\npython -m axolotl.cli.inference examples/your_config.yml --lora_model_dir=\"./lora-output-dir\"\nFull weights finetune:\npython -m axolotl.cli.inference examples/your_config.yml --base_model=\"./completed-model\"\nFull weights finetune w/ a prompt from a text file:\ncat /tmp/prompt.txt | python -m axolotl.cli.inference examples/your_config.yml \\\n  --base_model=\"./completed-model\" --prompter=None --load_in_8bit=True\n– With gradio hosting\npython -m axolotl.cli.inference examples/your_config.yml --gradio\n\nPlease use --sample_packing False if you have it on and receive the error similar to below:\n\nRuntimeError: stack expects each tensor to be equal size, but got [1, 32, 1, 128] at entry 0 and [1, 32, 8, 128] at entry 1\n\n\n\nMerge LORA to base\nThe following command will merge your LORA adapater with your base model. You can optionally pass the argument --lora_model_dir to specify the directory where your LORA adapter was saved, otherwhise, this will be inferred from output_dir in your axolotl config file. The merged model is saved in the sub-directory {lora_model_dir}/merged.\npython3 -m axolotl.cli.merge_lora your_config.yml --lora_model_dir=\"./completed-model\"\nYou may need to use the gpu_memory_limit and/or lora_on_cpu config options to avoid running out of memory. If you still run out of CUDA memory, you can try to merge in system RAM with\nCUDA_VISIBLE_DEVICES=\"\" python3 -m axolotl.cli.merge_lora ...\nalthough this will be very slow, and using the config options above are recommended instead.",
     "crumbs": [
       "Home"
     ]
@@ -731,7 +782,7 @@
     "href": "index.html#contributing",
     "title": "Axolotl",
     "section": "Contributing 🤝",
-    "text": "Contributing 🤝\nPlease read the contributing guide\nBugs? Please check the open issues else create a new Issue.\nPRs are greatly welcome!\nPlease run below to setup env\ngit clone https://github.com/OpenAccess-AI-Collective/axolotl\ncd axolotl\n\npip3 install packaging ninja\npip3 install -e '.[flash-attn,deepspeed]'\n\npip3 install -r requirements-dev.txt -r requirements-tests.txt\npre-commit install\n\n# test\npytest tests/\n\n# optional: run against all files\npre-commit run --all-files\nThanks to all of our contributors to date. Help drive open source AI progress forward by contributing to Axolotl.",
+    "text": "Contributing 🤝\nPlease read the contributing guide\nBugs? Please check the open issues else create a new Issue.\nPRs are greatly welcome!\nPlease run the quickstart instructions followed by the below to setup env:\npip3 install -r requirements-dev.txt -r requirements-tests.txt\npre-commit install\n\n# test\npytest tests/\n\n# optional: run against all files\npre-commit run --all-files\nThanks to all of our contributors to date. Help drive open source AI progress forward by contributing to Axolotl.",
     "crumbs": [
       "Home"
     ]
diff --git a/sitemap.xml b/sitemap.xml
index 6f379e90c..fbb04e17f 100644
--- a/sitemap.xml
+++ b/sitemap.xml
@@ -2,82 +2,86 @@
 <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/TODO.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/examples/colab-notebooks/colab-axolotl-example.html</loc>
-    <lastmod>2024-04-03T19:06:03.572Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.802Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/input_output.html</loc>
-    <lastmod>2024-04-03T19:06:03.572Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/debugging.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/nccl.html</loc>
-    <lastmod>2024-04-03T19:06:03.572Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/dataset-formats/template_free.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/dataset-formats/inst_tune.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/dataset-formats/conversation.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
-    <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/rlhf.html</loc>
-    <lastmod>2024-04-03T19:06:03.572Z</lastmod>
-  </url>
-  <url>
-    <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/faq.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
-  </url>
-  <url>
-    <loc>https://OpenAccess-AI-Collective.github.io/axolotl/FAQS.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/batch_vs_grad.html</loc>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/fsdp_qlora.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
+  </url>
+  <url>
+    <loc>https://OpenAccess-AI-Collective.github.io/axolotl/FAQS.html</loc>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
+  </url>
+  <url>
+    <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/faq.html</loc>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
+  </url>
+  <url>
+    <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/rlhf.html</loc>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/dataset-formats/pretraining.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/dataset-formats/tokenized.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/dataset-formats/index.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/mac.html</loc>
-    <lastmod>2024-04-03T19:06:03.572Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/multipack.html</loc>
-    <lastmod>2024-04-03T19:06:03.572Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/config.html</loc>
-    <lastmod>2024-04-03T19:06:03.568Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/docs/multi-node.html</loc>
-    <lastmod>2024-04-03T19:06:03.572Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.798Z</lastmod>
   </url>
   <url>
     <loc>https://OpenAccess-AI-Collective.github.io/axolotl/index.html</loc>
-    <lastmod>2024-04-03T19:06:03.580Z</lastmod>
+    <lastmod>2024-04-04T23:33:53.810Z</lastmod>
   </url>
 </urlset>
