diff --git a/.nojekyll b/.nojekyll
index 9c385668c..d7d8b5872 100644
--- a/.nojekyll
+++ b/.nojekyll
@@ -1 +1 @@
-c1261f78
\ No newline at end of file
+fd3551d4
\ No newline at end of file
diff --git a/docs/nd_parallelism.html b/docs/nd_parallelism.html
index 98c3adadf..932b9a150 100644
--- a/docs/nd_parallelism.html
+++ b/docs/nd_parallelism.html
@@ -7,7 +7,7 @@
 <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
 
-<title>nd_parallelism – Axolotl</title>
+<title>N-D Parallelism – Axolotl</title>
 <style>
 code{white-space: pre-wrap;}
 span.smallcaps{font-variant: small-caps;}
@@ -485,9 +485,7 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
     <h2 id="toc-title">On this page</h2>
    
   <ul>
-  <li><a href="#n-d-parallelism" id="toc-n-d-parallelism" class="nav-link active" data-scroll-target="#n-d-parallelism">N-D Parallelism</a>
-  <ul class="collapse">
-  <li><a href="#core-concepts" id="toc-core-concepts" class="nav-link" data-scroll-target="#core-concepts">Core Concepts</a>
+  <li><a href="#core-concepts" id="toc-core-concepts" class="nav-link active" data-scroll-target="#core-concepts">Core Concepts</a>
   <ul class="collapse">
   <li><a href="#sec-dp" id="toc-sec-dp" class="nav-link" data-scroll-target="#sec-dp">Data Parallelism</a></li>
   <li><a href="#sec-tp" id="toc-sec-tp" class="nav-link" data-scroll-target="#sec-tp">[Experimental] Tensor Parallelism (TP)</a></li>
@@ -497,18 +495,31 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
   <li><a href="#usage" id="toc-usage" class="nav-link" data-scroll-target="#usage">Usage</a></li>
   <li><a href="#examples" id="toc-examples" class="nav-link" data-scroll-target="#examples">Examples</a></li>
   <li><a href="#support-matrix" id="toc-support-matrix" class="nav-link" data-scroll-target="#support-matrix">Support Matrix</a></li>
-  </ul></li>
   </ul>
 </nav>
     </div>
 <!-- main -->
-<main class="content" id="quarto-document-content"><header id="title-block-header" class="quarto-title-block"><nav class="quarto-page-breadcrumbs quarto-title-breadcrumbs d-none d-lg-block" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="../docs/fsdp_qlora.html">Advanced Features</a></li><li class="breadcrumb-item"><a href="../docs/nd_parallelism.html">N-D Parallelism</a></li></ol></nav></header>
+<main class="content" id="quarto-document-content">
+
+<header id="title-block-header" class="quarto-title-block default"><nav class="quarto-page-breadcrumbs quarto-title-breadcrumbs d-none d-lg-block" aria-label="breadcrumb"><ol class="breadcrumb"><li class="breadcrumb-item"><a href="../docs/fsdp_qlora.html">Advanced Features</a></li><li class="breadcrumb-item"><a href="../docs/nd_parallelism.html">N-D Parallelism</a></li></ol></nav>
+<div class="quarto-title">
+<h1 class="title">N-D Parallelism</h1>
+</div>
 
 
 
+<div class="quarto-title-meta">
+
+    
+  
+    
+  </div>
+  
+
+
+</header>
+
 
-<section id="n-d-parallelism" class="level1">
-<h1>N-D Parallelism</h1>
 <p>Axolotl enables training models at scale by composing different parallelism techniques. This is essential when:</p>
 <ul>
 <li>A model’s weights are too large to fit on a single GPU’s memory.</li>
@@ -696,7 +707,6 @@ gtag('config', 'G-9KYCVJBNMQ', { 'anonymize_ip': true});
 </ul>
 
 
-</section>
 </section>
 
 </main> <!-- /main -->
diff --git a/search.json b/search.json
index d60443d17..07a1074d3 100644
--- a/search.json
+++ b/search.json
@@ -1599,7 +1599,7 @@
     "href": "docs/nd_parallelism.html",
     "title": "N-D Parallelism",
     "section": "",
-    "text": "Axolotl enables training models at scale by composing different parallelism techniques. This is essential when:\n\nA model’s weights are too large to fit on a single GPU’s memory.\nA model’s activations, especially with very long contexts, are too large for a single GPU.\nYou want to accelerate training by using multiple GPUs or nodes.\n\nor combinations of the above!\n\n\nParallelism strategies can be combined. The key is understanding how each one divides the workload. PyTorch’s DeviceMesh is the modern way to manage these combinations, creating a logical grid of your GPUs and assigning different parallel strategies to different dimensions of the grid.\n\n\nData Parallelism focuses on splitting the global data batch across GPUs.\n\nDistributed Data Parallel (DDP): The classic approach. The full model is replicated on every GPU. Each GPU processes a different slice of the data batch. Gradients are then averaged across all GPUs after the backward pass to keep the models synchronized. This can substantially improve data throughput compared to single-device training, but requires that each GPU is able to hold the entire model, its gradients, and optimizer states.\nFully Sharded Data Parallel (FSDP): A highly memory-efficient form of data parallelism (inspired by DeepSpeed’s ZeRO). Instead of replicating the model, FSDP shards the model’s parameters, gradients, and optimizer states across the GPUs in the data-parallel group. During computation, each GPU receives the specific parameters it needs via an all_gather operation just before they are used, and they can be discarded immediately after (reshard-after-forward).\n\nFSDP maps to ZeRO stages:\n\nZeRO-2 (reshard_after_forward=False): Shards gradients and optimizer states. Model weights are replicated on each GPU.\nZeRO-3 (reshard_after_forward=True): Shards gradients, optimizer states, AND model parameters. This provides the most memory savings at the cost of more communication (re-gathering parameters for both forward and backward passes).\n\n\n\n\n\n\nAlso known as “horizontal model parallelism,” as described in the Megatron-LM paper. Instead of splitting the batch, TP splits the model’s layers themselves across GPUs.\n\nHow it works: For a linear layer Y = XA, the weight matrix A is split column-wise (A = [A_1, A_2]). The computation becomes Y_1 = XA_1 and Y_2 = XA_2, which can happen in parallel on different GPUs. The final output Y is simply the concatenation of Y_1 and Y_2. Check this comment for more detailed info.\nRequirement: TP involves frequent, small communications within a forward/backward pass. It requires a very fast interconnect between GPUs (e.g., NVLink) and is typically not recommended across different nodes.\n\n\n\n\nContext Parallelism, also called Sequence Parallelism, addresses the memory bottleneck from long sequences. The input sequence itself is split along the sequence length dimension and distributed across GPUs.\n\nHow it works: If you have a sequence of 8192 tokens and a context_parallel_size of 4, each GPU will only handle a chunk of 2048 tokens.\nThe Challenge: Attention is not local; every token needs to “attend to” every other token. Splitting the sequence breaks this.\nThe Solution (ring-flash-attention): An efficient communication protocol is used. To compute attention for its local sequence chunk, each GPU passes its Key-Value (KV) cache to its neighbor in a “ring.” After N-1 steps, every GPU has seen the KV-cache from all other GPUs, allowing it to compute the correct attention values for its chunk. This is implemented using the highly optimized flash-attention kernel at each step.\n\n\n\n\nHSDP is a 2D strategy that intelligently combines FSDP and DDP, typically for multi-node training.\n\nIntra-Node (within a machine): Use FSDP. This is efficient because GPUs on the same node have fast interconnects (NVLink), making the all_gather operations for sharded parameters fast.\nInter-Node (across machines): Use DDP. The gradient synchronization between nodes is less frequent than FSDP’s parameter gathering, making it a better fit for the slower node-to-node network (e.g., Ethernet/Infiniband).\nExample: With 2 nodes of 8 GPUs each (16 total), you could have dp_shard_size=8 (FSDP within each node) and dp_replicate_size=2 (DDP across the two nodes).\n\n\n\n\n\n# FSDP config. See https://docs.axolotl.ai/docs/multi-gpu.html#sec-fsdp\nfsdp_version: 2\nfsdp_config:\n  # ...\n\n# The number of GPUs to shard the model parameters across (FSDP dimension).\ndp_shard_size: 4\n\n# The number of times to replicate the sharded model (DDP dimension).\ndp_replicate_size: 2\n\n# Number of GPUs for Tensor Parallelism.\ntensor_parallel_size: 1  # (default is 1, no TP)\n\n# Number of GPUs for Context/Sequence Parallelism.\ncontext_parallel_size: 1 # (default is 1, no CP)\nNote: We recommend FSDP. DeepSpeed is only compatible with tensor_parallel_size.\n\n\n\n\nHSDP on 2 nodes with 4 GPUs each (8 GPUs total):\n\nYou want FSDP within each node and DDP across nodes.\nSet dp_shard_size: 4 and dp_replicate_size: 2.\n\nFSDP + TP on a single 8-GPU node:\n\nYou want to split the model across 4 GPUs using FSDP, and further split each layer across 2 GPUs with TP.\nSet dp_shard_size: 4 and tensor_parallel_size: 2.\n\nFSDP + CP on a single 8-GPU node for long context:\n\nYou want to shard the model across all 8 GPUs and also split the sequence length across all 8 GPUs.\nSet dp_shard_size: 8 and context_parallel_size: 8. Note: this means the data parallel group and context parallel group are the same. A more common setup might be to shard across a smaller group.\n\n\n\n\n\nThis matrix describes how different parallelism methods can be combined in Axolotl.\n\n\n\n\n\n\n\n\n\n\n\nCombination\ndp_replicate_size\ndp_shard_size\ntp_size\ncp_size\nStatus & Notes\n\n\n\n\nFSDP (ZeRO-3)\n1\n&gt;1\n1\n1\n✅ Fully supported. Shards model across all GPUs.\n\n\nHSDP\n&gt;1\n&gt;1\n1\n1\n✅ Fully supported. FSDP intra-node, DDP inter-node.\n\n\nFSDP + TP\n1\n&gt;1\n&gt;1\n1\n✅ 2D Parallelism. Shards the model across a dp_shard group, and TP-splits layers within the tp group.\n\n\nHSDP + TP\n&gt;1\n&gt;1\n&gt;1\n1\n✅ 3D Parallelism. A powerful but complex combination.\n\n\nFSDP + CP\n1\n&gt;1\n1\n&gt;1\n✅ 2D Parallelism. Combines FSDP with context parallelism.\n\n\nFSDP + TP + CP\n1\n&gt;1\n&gt;1\n&gt;1\n✅ 3D Parallelism. Another advanced combination.\n\n\nDDP + TP/CP\n&gt;1\n1\n&gt;1\n&gt;1\n❌ Not Supported. The ParallelismConfig explicitly prevents this, as composing pure DDP with TP/CP without FSDP is inefficient and complex. You should use FSDP instead (dp_shard_size &gt; 1).\n\n\nJust TP / CP\n1\n1\n&gt;1\n&gt;1\n✅ Supported. Useful for inference or when the model fits on one GPU but context is too long.\n\n\n\n\ntp_size refers to tensor_parallel_size\ncp_size refers to context_parallel_size",
+    "text": "Axolotl enables training models at scale by composing different parallelism techniques. This is essential when:\nor combinations of the above!",
     "crumbs": [
       "Advanced Features",
       "N-D Parallelism"
@@ -1609,8 +1609,8 @@
     "objectID": "docs/nd_parallelism.html#core-concepts",
     "href": "docs/nd_parallelism.html#core-concepts",
     "title": "N-D Parallelism",
-    "section": "",
-    "text": "Parallelism strategies can be combined. The key is understanding how each one divides the workload. PyTorch’s DeviceMesh is the modern way to manage these combinations, creating a logical grid of your GPUs and assigning different parallel strategies to different dimensions of the grid.\n\n\nData Parallelism focuses on splitting the global data batch across GPUs.\n\nDistributed Data Parallel (DDP): The classic approach. The full model is replicated on every GPU. Each GPU processes a different slice of the data batch. Gradients are then averaged across all GPUs after the backward pass to keep the models synchronized. This can substantially improve data throughput compared to single-device training, but requires that each GPU is able to hold the entire model, its gradients, and optimizer states.\nFully Sharded Data Parallel (FSDP): A highly memory-efficient form of data parallelism (inspired by DeepSpeed’s ZeRO). Instead of replicating the model, FSDP shards the model’s parameters, gradients, and optimizer states across the GPUs in the data-parallel group. During computation, each GPU receives the specific parameters it needs via an all_gather operation just before they are used, and they can be discarded immediately after (reshard-after-forward).\n\nFSDP maps to ZeRO stages:\n\nZeRO-2 (reshard_after_forward=False): Shards gradients and optimizer states. Model weights are replicated on each GPU.\nZeRO-3 (reshard_after_forward=True): Shards gradients, optimizer states, AND model parameters. This provides the most memory savings at the cost of more communication (re-gathering parameters for both forward and backward passes).\n\n\n\n\n\n\nAlso known as “horizontal model parallelism,” as described in the Megatron-LM paper. Instead of splitting the batch, TP splits the model’s layers themselves across GPUs.\n\nHow it works: For a linear layer Y = XA, the weight matrix A is split column-wise (A = [A_1, A_2]). The computation becomes Y_1 = XA_1 and Y_2 = XA_2, which can happen in parallel on different GPUs. The final output Y is simply the concatenation of Y_1 and Y_2. Check this comment for more detailed info.\nRequirement: TP involves frequent, small communications within a forward/backward pass. It requires a very fast interconnect between GPUs (e.g., NVLink) and is typically not recommended across different nodes.\n\n\n\n\nContext Parallelism, also called Sequence Parallelism, addresses the memory bottleneck from long sequences. The input sequence itself is split along the sequence length dimension and distributed across GPUs.\n\nHow it works: If you have a sequence of 8192 tokens and a context_parallel_size of 4, each GPU will only handle a chunk of 2048 tokens.\nThe Challenge: Attention is not local; every token needs to “attend to” every other token. Splitting the sequence breaks this.\nThe Solution (ring-flash-attention): An efficient communication protocol is used. To compute attention for its local sequence chunk, each GPU passes its Key-Value (KV) cache to its neighbor in a “ring.” After N-1 steps, every GPU has seen the KV-cache from all other GPUs, allowing it to compute the correct attention values for its chunk. This is implemented using the highly optimized flash-attention kernel at each step.\n\n\n\n\nHSDP is a 2D strategy that intelligently combines FSDP and DDP, typically for multi-node training.\n\nIntra-Node (within a machine): Use FSDP. This is efficient because GPUs on the same node have fast interconnects (NVLink), making the all_gather operations for sharded parameters fast.\nInter-Node (across machines): Use DDP. The gradient synchronization between nodes is less frequent than FSDP’s parameter gathering, making it a better fit for the slower node-to-node network (e.g., Ethernet/Infiniband).\nExample: With 2 nodes of 8 GPUs each (16 total), you could have dp_shard_size=8 (FSDP within each node) and dp_replicate_size=2 (DDP across the two nodes).",
+    "section": "Core Concepts",
+    "text": "Core Concepts\nParallelism strategies can be combined. The key is understanding how each one divides the workload. PyTorch’s DeviceMesh is the modern way to manage these combinations, creating a logical grid of your GPUs and assigning different parallel strategies to different dimensions of the grid.\n\nData Parallelism\nData Parallelism focuses on splitting the global data batch across GPUs.\n\nDistributed Data Parallel (DDP): The classic approach. The full model is replicated on every GPU. Each GPU processes a different slice of the data batch. Gradients are then averaged across all GPUs after the backward pass to keep the models synchronized. This can substantially improve data throughput compared to single-device training, but requires that each GPU is able to hold the entire model, its gradients, and optimizer states.\nFully Sharded Data Parallel (FSDP): A highly memory-efficient form of data parallelism (inspired by DeepSpeed’s ZeRO). Instead of replicating the model, FSDP shards the model’s parameters, gradients, and optimizer states across the GPUs in the data-parallel group. During computation, each GPU receives the specific parameters it needs via an all_gather operation just before they are used, and they can be discarded immediately after (reshard-after-forward).\n\nFSDP maps to ZeRO stages:\n\nZeRO-2 (reshard_after_forward=False): Shards gradients and optimizer states. Model weights are replicated on each GPU.\nZeRO-3 (reshard_after_forward=True): Shards gradients, optimizer states, AND model parameters. This provides the most memory savings at the cost of more communication (re-gathering parameters for both forward and backward passes).\n\n\n\n\n\n[Experimental] Tensor Parallelism (TP)\nAlso known as “horizontal model parallelism,” as described in the Megatron-LM paper. Instead of splitting the batch, TP splits the model’s layers themselves across GPUs.\n\nHow it works: For a linear layer Y = XA, the weight matrix A is split column-wise (A = [A_1, A_2]). The computation becomes Y_1 = XA_1 and Y_2 = XA_2, which can happen in parallel on different GPUs. The final output Y is simply the concatenation of Y_1 and Y_2. Check this comment for more detailed info.\nRequirement: TP involves frequent, small communications within a forward/backward pass. It requires a very fast interconnect between GPUs (e.g., NVLink) and is typically not recommended across different nodes.\n\n\n\nContext Parallelism (CP)\nContext Parallelism, also called Sequence Parallelism, addresses the memory bottleneck from long sequences. The input sequence itself is split along the sequence length dimension and distributed across GPUs.\n\nHow it works: If you have a sequence of 8192 tokens and a context_parallel_size of 4, each GPU will only handle a chunk of 2048 tokens.\nThe Challenge: Attention is not local; every token needs to “attend to” every other token. Splitting the sequence breaks this.\nThe Solution (ring-flash-attention): An efficient communication protocol is used. To compute attention for its local sequence chunk, each GPU passes its Key-Value (KV) cache to its neighbor in a “ring.” After N-1 steps, every GPU has seen the KV-cache from all other GPUs, allowing it to compute the correct attention values for its chunk. This is implemented using the highly optimized flash-attention kernel at each step.\n\n\n\nHybrid Sharding Data Parallel (HSDP)\nHSDP is a 2D strategy that intelligently combines FSDP and DDP, typically for multi-node training.\n\nIntra-Node (within a machine): Use FSDP. This is efficient because GPUs on the same node have fast interconnects (NVLink), making the all_gather operations for sharded parameters fast.\nInter-Node (across machines): Use DDP. The gradient synchronization between nodes is less frequent than FSDP’s parameter gathering, making it a better fit for the slower node-to-node network (e.g., Ethernet/Infiniband).\nExample: With 2 nodes of 8 GPUs each (16 total), you could have dp_shard_size=8 (FSDP within each node) and dp_replicate_size=2 (DDP across the two nodes).",
     "crumbs": [
       "Advanced Features",
       "N-D Parallelism"
@@ -1620,8 +1620,8 @@
     "objectID": "docs/nd_parallelism.html#usage",
     "href": "docs/nd_parallelism.html#usage",
     "title": "N-D Parallelism",
-    "section": "",
-    "text": "# FSDP config. See https://docs.axolotl.ai/docs/multi-gpu.html#sec-fsdp\nfsdp_version: 2\nfsdp_config:\n  # ...\n\n# The number of GPUs to shard the model parameters across (FSDP dimension).\ndp_shard_size: 4\n\n# The number of times to replicate the sharded model (DDP dimension).\ndp_replicate_size: 2\n\n# Number of GPUs for Tensor Parallelism.\ntensor_parallel_size: 1  # (default is 1, no TP)\n\n# Number of GPUs for Context/Sequence Parallelism.\ncontext_parallel_size: 1 # (default is 1, no CP)\nNote: We recommend FSDP. DeepSpeed is only compatible with tensor_parallel_size.",
+    "section": "Usage",
+    "text": "Usage\n# FSDP config. See https://docs.axolotl.ai/docs/multi-gpu.html#sec-fsdp\nfsdp_version: 2\nfsdp_config:\n  # ...\n\n# The number of GPUs to shard the model parameters across (FSDP dimension).\ndp_shard_size: 4\n\n# The number of times to replicate the sharded model (DDP dimension).\ndp_replicate_size: 2\n\n# Number of GPUs for Tensor Parallelism.\ntensor_parallel_size: 1  # (default is 1, no TP)\n\n# Number of GPUs for Context/Sequence Parallelism.\ncontext_parallel_size: 1 # (default is 1, no CP)\nNote: We recommend FSDP. DeepSpeed is only compatible with tensor_parallel_size.",
     "crumbs": [
       "Advanced Features",
       "N-D Parallelism"
@@ -1631,8 +1631,8 @@
     "objectID": "docs/nd_parallelism.html#examples",
     "href": "docs/nd_parallelism.html#examples",
     "title": "N-D Parallelism",
-    "section": "",
-    "text": "HSDP on 2 nodes with 4 GPUs each (8 GPUs total):\n\nYou want FSDP within each node and DDP across nodes.\nSet dp_shard_size: 4 and dp_replicate_size: 2.\n\nFSDP + TP on a single 8-GPU node:\n\nYou want to split the model across 4 GPUs using FSDP, and further split each layer across 2 GPUs with TP.\nSet dp_shard_size: 4 and tensor_parallel_size: 2.\n\nFSDP + CP on a single 8-GPU node for long context:\n\nYou want to shard the model across all 8 GPUs and also split the sequence length across all 8 GPUs.\nSet dp_shard_size: 8 and context_parallel_size: 8. Note: this means the data parallel group and context parallel group are the same. A more common setup might be to shard across a smaller group.",
+    "section": "Examples",
+    "text": "Examples\n\nHSDP on 2 nodes with 4 GPUs each (8 GPUs total):\n\nYou want FSDP within each node and DDP across nodes.\nSet dp_shard_size: 4 and dp_replicate_size: 2.\n\nFSDP + TP on a single 8-GPU node:\n\nYou want to split the model across 4 GPUs using FSDP, and further split each layer across 2 GPUs with TP.\nSet dp_shard_size: 4 and tensor_parallel_size: 2.\n\nFSDP + CP on a single 8-GPU node for long context:\n\nYou want to shard the model across all 8 GPUs and also split the sequence length across all 8 GPUs.\nSet dp_shard_size: 8 and context_parallel_size: 8. Note: this means the data parallel group and context parallel group are the same. A more common setup might be to shard across a smaller group.",
     "crumbs": [
       "Advanced Features",
       "N-D Parallelism"
@@ -1642,8 +1642,8 @@
     "objectID": "docs/nd_parallelism.html#support-matrix",
     "href": "docs/nd_parallelism.html#support-matrix",
     "title": "N-D Parallelism",
-    "section": "",
-    "text": "This matrix describes how different parallelism methods can be combined in Axolotl.\n\n\n\n\n\n\n\n\n\n\n\nCombination\ndp_replicate_size\ndp_shard_size\ntp_size\ncp_size\nStatus & Notes\n\n\n\n\nFSDP (ZeRO-3)\n1\n&gt;1\n1\n1\n✅ Fully supported. Shards model across all GPUs.\n\n\nHSDP\n&gt;1\n&gt;1\n1\n1\n✅ Fully supported. FSDP intra-node, DDP inter-node.\n\n\nFSDP + TP\n1\n&gt;1\n&gt;1\n1\n✅ 2D Parallelism. Shards the model across a dp_shard group, and TP-splits layers within the tp group.\n\n\nHSDP + TP\n&gt;1\n&gt;1\n&gt;1\n1\n✅ 3D Parallelism. A powerful but complex combination.\n\n\nFSDP + CP\n1\n&gt;1\n1\n&gt;1\n✅ 2D Parallelism. Combines FSDP with context parallelism.\n\n\nFSDP + TP + CP\n1\n&gt;1\n&gt;1\n&gt;1\n✅ 3D Parallelism. Another advanced combination.\n\n\nDDP + TP/CP\n&gt;1\n1\n&gt;1\n&gt;1\n❌ Not Supported. The ParallelismConfig explicitly prevents this, as composing pure DDP with TP/CP without FSDP is inefficient and complex. You should use FSDP instead (dp_shard_size &gt; 1).\n\n\nJust TP / CP\n1\n1\n&gt;1\n&gt;1\n✅ Supported. Useful for inference or when the model fits on one GPU but context is too long.\n\n\n\n\ntp_size refers to tensor_parallel_size\ncp_size refers to context_parallel_size",
+    "section": "Support Matrix",
+    "text": "Support Matrix\nThis matrix describes how different parallelism methods can be combined in Axolotl.\n\n\n\n\n\n\n\n\n\n\n\nCombination\ndp_replicate_size\ndp_shard_size\ntp_size\ncp_size\nStatus & Notes\n\n\n\n\nFSDP (ZeRO-3)\n1\n&gt;1\n1\n1\n✅ Fully supported. Shards model across all GPUs.\n\n\nHSDP\n&gt;1\n&gt;1\n1\n1\n✅ Fully supported. FSDP intra-node, DDP inter-node.\n\n\nFSDP + TP\n1\n&gt;1\n&gt;1\n1\n✅ 2D Parallelism. Shards the model across a dp_shard group, and TP-splits layers within the tp group.\n\n\nHSDP + TP\n&gt;1\n&gt;1\n&gt;1\n1\n✅ 3D Parallelism. A powerful but complex combination.\n\n\nFSDP + CP\n1\n&gt;1\n1\n&gt;1\n✅ 2D Parallelism. Combines FSDP with context parallelism.\n\n\nFSDP + TP + CP\n1\n&gt;1\n&gt;1\n&gt;1\n✅ 3D Parallelism. Another advanced combination.\n\n\nDDP + TP/CP\n&gt;1\n1\n&gt;1\n&gt;1\n❌ Not Supported. The ParallelismConfig explicitly prevents this, as composing pure DDP with TP/CP without FSDP is inefficient and complex. You should use FSDP instead (dp_shard_size &gt; 1).\n\n\nJust TP / CP\n1\n1\n&gt;1\n&gt;1\n✅ Supported. Useful for inference or when the model fits on one GPU but context is too long.\n\n\n\n\ntp_size refers to tensor_parallel_size\ncp_size refers to context_parallel_size",
     "crumbs": [
       "Advanced Features",
       "N-D Parallelism"
diff --git a/sitemap.xml b/sitemap.xml
index 80b4489a3..696b1ba22 100644
--- a/sitemap.xml
+++ b/sitemap.xml
@@ -2,790 +2,790 @@
 <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
   <url>
     <loc>https://docs.axolotl.ai/TODO.html</loc>
-    <lastmod>2025-08-04T00:05:24.909Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.770Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/index.html</loc>
-    <lastmod>2025-08-04T00:05:24.930Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.791Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/debugging.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.772Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/amd_hpc.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.callbacks.mlflow_.html</loc>
-    <lastmod>2025-08-04T00:09:06.121Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.385Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.llama_expand_mask.html</loc>
-    <lastmod>2025-08-04T00:09:05.543Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.787Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/loaders.patch_manager.html</loc>
-    <lastmod>2025-08-04T00:09:05.152Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.383Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.chat.format.llama3x.html</loc>
-    <lastmod>2025-08-04T00:09:04.830Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.053Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.train.html</loc>
-    <lastmod>2025-08-04T00:09:04.887Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.112Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.callbacks.perplexity.html</loc>
-    <lastmod>2025-08-04T00:09:06.112Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.376Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.chat.messages.html</loc>
-    <lastmod>2025-08-04T00:09:04.827Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.050Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.callbacks.lisa.html</loc>
-    <lastmod>2025-08-04T00:09:06.118Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.381Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.merge_sharded_fsdp_weights.html</loc>
-    <lastmod>2025-08-04T00:09:04.983Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.210Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.mixtral.html</loc>
-    <lastmod>2025-08-04T00:09:05.603Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.850Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.chat_templates.html</loc>
-    <lastmod>2025-08-04T00:09:05.641Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.889Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.chat.format.shared.html</loc>
-    <lastmod>2025-08-04T00:09:04.831Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.055Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.mixins.optimizer.html</loc>
-    <lastmod>2025-08-04T00:09:05.159Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.390Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.collators.mamba.html</loc>
-    <lastmod>2025-08-04T00:09:06.060Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.322Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/logging_config.html</loc>
-    <lastmod>2025-08-04T00:09:04.775Z</lastmod>
+    <lastmod>2025-08-04T14:27:13.999Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.collators.mm_chat.html</loc>
-    <lastmod>2025-08-04T00:09:06.065Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.327Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.completion.html</loc>
-    <lastmod>2025-08-04T00:09:05.281Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.515Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/kernels.utils.html</loc>
-    <lastmod>2025-08-04T00:09:05.495Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.737Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.chat_template.html</loc>
-    <lastmod>2025-08-04T00:09:05.314Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.551Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/kernels.swiglu.html</loc>
-    <lastmod>2025-08-04T00:09:05.486Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.728Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/common.const.html</loc>
-    <lastmod>2025-08-04T00:09:06.021Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.281Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.cloud.base.html</loc>
-    <lastmod>2025-08-04T00:09:05.007Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.234Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.orpo.chat_template.html</loc>
-    <lastmod>2025-08-04T00:09:05.378Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.616Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.builders.rl.html</loc>
-    <lastmod>2025-08-04T00:09:04.791Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.014Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.dict.html</loc>
-    <lastmod>2025-08-04T00:09:05.733Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.984Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.integrations.html</loc>
-    <lastmod>2025-08-04T00:09:05.844Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.099Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.utils.html</loc>
-    <lastmod>2025-08-04T00:09:05.117Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.346Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.trainer_fsdp_optim.html</loc>
-    <lastmod>2025-08-04T00:09:05.592Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.838Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.evaluate.html</loc>
-    <lastmod>2025-08-04T00:09:04.895Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.121Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.builders.causal.html</loc>
-    <lastmod>2025-08-04T00:09:04.786Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.010Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.multipack.html</loc>
-    <lastmod>2025-08-04T00:09:05.538Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.782Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.llama_patch_multipack.html</loc>
-    <lastmod>2025-08-04T00:09:05.583Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.829Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.delinearize_llama4.html</loc>
-    <lastmod>2025-08-04T00:09:04.948Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.174Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.trl.html</loc>
-    <lastmod>2025-08-04T00:09:05.827Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.081Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.zephyr.html</loc>
-    <lastmod>2025-08-04T00:09:05.336Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.574Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/integrations.kd.trainer.html</loc>
-    <lastmod>2025-08-04T00:09:06.008Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.268Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.gradient_checkpointing.offload_disk.html</loc>
-    <lastmod>2025-08-04T00:09:05.632Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.880Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.optimizers.adopt.html</loc>
-    <lastmod>2025-08-04T00:09:05.740Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.992Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.data.batch_dataset_fetcher.html</loc>
-    <lastmod>2025-08-04T00:09:05.602Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.849Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.cloud.modal_.html</loc>
-    <lastmod>2025-08-04T00:09:05.014Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.240Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_chat.html</loc>
-    <lastmod>2025-08-04T00:09:05.240Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.474Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.freeze.html</loc>
-    <lastmod>2025-08-04T00:09:05.663Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.912Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.bradley_terry.llama3.html</loc>
-    <lastmod>2025-08-04T00:09:05.381Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.620Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/integrations.base.html</loc>
-    <lastmod>2025-08-04T00:09:05.996Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.255Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.unsloth_.html</loc>
-    <lastmod>2025-08-04T00:09:05.600Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.847Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.kto.chatml.html</loc>
-    <lastmod>2025-08-04T00:09:05.356Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.594Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.main.html</loc>
-    <lastmod>2025-08-04T00:09:04.878Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.104Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/common.datasets.html</loc>
-    <lastmod>2025-08-04T00:09:06.036Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.297Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/train.html</loc>
-    <lastmod>2025-08-04T00:09:04.689Z</lastmod>
+    <lastmod>2025-08-04T14:27:13.911Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.trainer.html</loc>
-    <lastmod>2025-08-04T00:09:05.680Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.930Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.llama2_chat.html</loc>
-    <lastmod>2025-08-04T00:09:05.275Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.509Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/index.html</loc>
-    <lastmod>2025-08-04T00:09:04.627Z</lastmod>
+    <lastmod>2025-08-04T14:27:13.849Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.chat_template.html</loc>
-    <lastmod>2025-08-04T00:09:05.227Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.460Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.training_args.html</loc>
-    <lastmod>2025-08-04T00:09:04.803Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.027Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/kernels.quantize.html</loc>
-    <lastmod>2025-08-04T00:09:05.494Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.735Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/convert.html</loc>
-    <lastmod>2025-08-04T00:09:04.724Z</lastmod>
+    <lastmod>2025-08-04T14:27:13.947Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/integrations.grokfast.optimizer.html</loc>
-    <lastmod>2025-08-04T00:09:06.000Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.260Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.stepwise_supervised.html</loc>
-    <lastmod>2025-08-04T00:09:05.291Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.526Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.model.html</loc>
-    <lastmod>2025-08-04T00:09:05.790Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.043Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.callbacks.qat.html</loc>
-    <lastmod>2025-08-04T00:09:06.132Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.396Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/loaders.constants.html</loc>
-    <lastmod>2025-08-04T00:09:05.154Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.384Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.utils.sweeps.html</loc>
-    <lastmod>2025-08-04T00:09:05.044Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.271Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.llama3.html</loc>
-    <lastmod>2025-08-04T00:09:05.325Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.562Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.datasets.transforms.chat_builder.html</loc>
-    <lastmod>2025-08-04T00:09:04.844Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.068Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.utils.fetch.html</loc>
-    <lastmod>2025-08-04T00:09:05.032Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.259Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.mamba.html</loc>
-    <lastmod>2025-08-04T00:09:05.086Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.314Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.enums.html</loc>
-    <lastmod>2025-08-04T00:09:05.855Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.109Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.callbacks.profiler.html</loc>
-    <lastmod>2025-08-04T00:09:06.116Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.380Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.metharme.html</loc>
-    <lastmod>2025-08-04T00:09:05.298Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.534Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.trl.html</loc>
-    <lastmod>2025-08-04T00:09:05.080Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.308Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.orcamini.html</loc>
-    <lastmod>2025-08-04T00:09:05.302Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.538Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.samplers.multipack.html</loc>
-    <lastmod>2025-08-04T00:09:06.106Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.369Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schedulers.html</loc>
-    <lastmod>2025-08-04T00:09:05.707Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.958Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.grpo.trainer.html</loc>
-    <lastmod>2025-08-04T00:09:05.104Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.331Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_tokenizers.html</loc>
-    <lastmod>2025-08-04T00:09:04.765Z</lastmod>
+    <lastmod>2025-08-04T14:27:13.989Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/config-reference.html</loc>
-    <lastmod>2025-08-04T00:09:18.528Z</lastmod>
+    <lastmod>2025-08-04T14:27:28.691Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/multimodal.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/mixed_precision.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/unsloth.html</loc>
-    <lastmod>2025-08-04T00:05:24.915Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.776Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/ray-integration.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset-formats/stepwise_supervised.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset-formats/template_free.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset-formats/index.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset-formats/pretraining.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/nd_parallelism.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/sequence_parallelism.html</loc>
-    <lastmod>2025-08-04T00:05:24.915Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.776Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/inference.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/fsdp_qlora.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.772Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/multi-node.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/lora_optims.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/getting-started.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.772Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset_loading.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/lr_groups.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/mac.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/src/axolotl/integrations/cut_cross_entropy/ACKNOWLEDGEMENTS.html</loc>
-    <lastmod>2025-08-04T00:05:24.934Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.795Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/src/axolotl/integrations/LICENSE.html</loc>
-    <lastmod>2025-08-04T00:05:24.934Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.795Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/input_output.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/gradient_checkpointing.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.772Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/qat.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/faq.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.772Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset_preprocessing.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.772Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/nccl.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/cli.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/torchao.html</loc>
-    <lastmod>2025-08-04T00:05:24.915Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.776Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/multi-gpu.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/rlhf.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset-formats/tokenized.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset-formats/conversation.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/dataset-formats/inst_tune.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/reward_modelling.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/docker.html</loc>
-    <lastmod>2025-08-04T00:05:24.911Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.772Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/installation.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/quantize.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/custom_integrations.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/batch_vs_grad.html</loc>
-    <lastmod>2025-08-04T00:05:24.910Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.771Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.utils.train.html</loc>
-    <lastmod>2025-08-04T00:09:05.054Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.282Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.art.html</loc>
-    <lastmod>2025-08-04T00:09:04.918Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.143Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.grpo.sampler.html</loc>
-    <lastmod>2025-08-04T00:09:05.116Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.344Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/loaders.model.html</loc>
-    <lastmod>2025-08-04T00:09:05.127Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.356Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.preprocess.html</loc>
-    <lastmod>2025-08-04T00:09:04.992Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.218Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.utils.html</loc>
-    <lastmod>2025-08-04T00:09:05.015Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.242Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.inference.html</loc>
-    <lastmod>2025-08-04T00:09:04.962Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.189Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.btlm_attn_hijack_flash.html</loc>
-    <lastmod>2025-08-04T00:09:05.582Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.827Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/datasets.html</loc>
-    <lastmod>2025-08-04T00:09:04.710Z</lastmod>
+    <lastmod>2025-08-04T14:27:13.933Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.transformers_fa_utils.html</loc>
-    <lastmod>2025-08-04T00:09:05.599Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.845Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.llama_attn_hijack_flash.html</loc>
-    <lastmod>2025-08-04T00:09:05.521Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.764Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.relora.html</loc>
-    <lastmod>2025-08-04T00:09:05.542Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.786Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.stablelm_attn_hijack_flash.html</loc>
-    <lastmod>2025-08-04T00:09:05.589Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.835Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/loaders.adapter.html</loc>
-    <lastmod>2025-08-04T00:09:05.142Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.372Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.dpo.trainer.html</loc>
-    <lastmod>2025-08-04T00:09:05.093Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.320Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/integrations.cut_cross_entropy.args.html</loc>
-    <lastmod>2025-08-04T00:09:05.999Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.259Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.utils.html</loc>
-    <lastmod>2025-08-04T00:09:05.580Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.826Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/loaders.processor.html</loc>
-    <lastmod>2025-08-04T00:09:05.137Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.366Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.config.html</loc>
-    <lastmod>2025-08-04T00:09:04.943Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.169Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/integrations.liger.args.html</loc>
-    <lastmod>2025-08-04T00:09:06.011Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.271Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/loaders.tokenizer.html</loc>
-    <lastmod>2025-08-04T00:09:05.135Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.365Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.config.html</loc>
-    <lastmod>2025-08-04T00:09:05.783Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.036Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.ctx_managers.sequence_parallel.html</loc>
-    <lastmod>2025-08-04T00:09:05.193Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.425Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.mixins.scheduler.html</loc>
-    <lastmod>2025-08-04T00:09:05.169Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.400Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.base.html</loc>
-    <lastmod>2025-08-04T00:09:05.065Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.293Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.utils.args.html</loc>
-    <lastmod>2025-08-04T00:09:05.027Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.253Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.messages.chat.html</loc>
-    <lastmod>2025-08-04T00:09:05.313Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.549Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.lora_kernels.html</loc>
-    <lastmod>2025-08-04T00:09:05.572Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.816Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/kernels.lora.html</loc>
-    <lastmod>2025-08-04T00:09:05.465Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.706Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.vllm_serve.html</loc>
-    <lastmod>2025-08-04T00:09:05.004Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.230Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.multimodal.html</loc>
-    <lastmod>2025-08-04T00:09:05.832Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.086Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.utils.html</loc>
-    <lastmod>2025-08-04T00:09:05.860Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.115Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.llama_attn_hijack_xformers.html</loc>
-    <lastmod>2025-08-04T00:09:05.523Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.765Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/integrations.lm_eval.args.html</loc>
-    <lastmod>2025-08-04T00:09:06.014Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.275Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.mistral_attn_hijack_flash.html</loc>
-    <lastmod>2025-08-04T00:09:05.537Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.781Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.collators.core.html</loc>
-    <lastmod>2025-08-04T00:09:06.038Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.299Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.chat.format.chatml.html</loc>
-    <lastmod>2025-08-04T00:09:04.828Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.052Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.passthrough.html</loc>
-    <lastmod>2025-08-04T00:09:05.339Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.577Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.datasets.chat.html</loc>
-    <lastmod>2025-08-04T00:09:04.836Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.060Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.bench.html</loc>
-    <lastmod>2025-08-04T00:09:05.655Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.904Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.training.html</loc>
-    <lastmod>2025-08-04T00:09:05.797Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.050Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.collators.batching.html</loc>
-    <lastmod>2025-08-04T00:09:06.057Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.318Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.input_output.html</loc>
-    <lastmod>2025-08-04T00:09:05.287Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.522Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.lora.html</loc>
-    <lastmod>2025-08-04T00:09:05.646Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.894Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.base.html</loc>
-    <lastmod>2025-08-04T00:09:05.194Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.427Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_w_system.html</loc>
-    <lastmod>2025-08-04T00:09:05.254Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.488Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.datasets.html</loc>
-    <lastmod>2025-08-04T00:09:05.815Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.068Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.user_defined.html</loc>
-    <lastmod>2025-08-04T00:09:05.338Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.576Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.schemas.peft.html</loc>
-    <lastmod>2025-08-04T00:09:05.823Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.077Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.pygmalion.html</loc>
-    <lastmod>2025-08-04T00:09:05.308Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.545Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/common.architectures.html</loc>
-    <lastmod>2025-08-04T00:09:06.019Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.280Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/monkeypatch.gradient_checkpointing.offload_cpu.html</loc>
-    <lastmod>2025-08-04T00:09:05.607Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.854Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.callbacks.comet_.html</loc>
-    <lastmod>2025-08-04T00:09:06.125Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.389Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/integrations.spectrum.args.html</loc>
-    <lastmod>2025-08-04T00:09:06.018Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.278Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.quantize.html</loc>
-    <lastmod>2025-08-04T00:09:04.997Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.223Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.checks.html</loc>
-    <lastmod>2025-08-04T00:09:04.925Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.151Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.kto.llama3.html</loc>
-    <lastmod>2025-08-04T00:09:05.347Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.585Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.model_shard_quant.html</loc>
-    <lastmod>2025-08-04T00:09:05.651Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.900Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.quantization.html</loc>
-    <lastmod>2025-08-04T00:09:05.769Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.022Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.trainers.mixins.rng_state_loader.html</loc>
-    <lastmod>2025-08-04T00:09:05.163Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.393Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/kernels.geglu.html</loc>
-    <lastmod>2025-08-04T00:09:05.476Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.717Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.data.pretraining.html</loc>
-    <lastmod>2025-08-04T00:09:05.742Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.994Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.kto.user_defined.html</loc>
-    <lastmod>2025-08-04T00:09:05.357Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.595Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/core.builders.base.html</loc>
-    <lastmod>2025-08-04T00:09:04.781Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.005Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.merge_lora.html</loc>
-    <lastmod>2025-08-04T00:09:04.971Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.198Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.utils.load.html</loc>
-    <lastmod>2025-08-04T00:09:05.038Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.265Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.data.sft.html</loc>
-    <lastmod>2025-08-04T00:09:05.748Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.001Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.user_defined.html</loc>
-    <lastmod>2025-08-04T00:09:05.262Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.496Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.tokenization.html</loc>
-    <lastmod>2025-08-04T00:09:05.639Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.888Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.dpo.chatml.html</loc>
-    <lastmod>2025-08-04T00:09:05.335Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.573Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/models.mamba.modeling_mamba.html</loc>
-    <lastmod>2025-08-04T00:09:06.037Z</lastmod>
+    <lastmod>2025-08-04T14:27:15.298Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/cli.args.html</loc>
-    <lastmod>2025-08-04T00:09:04.914Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.140Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/evaluate.html</loc>
-    <lastmod>2025-08-04T00:09:04.699Z</lastmod>
+    <lastmod>2025-08-04T14:27:13.922Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/prompt_strategies.alpaca_instruct.html</loc>
-    <lastmod>2025-08-04T00:09:05.242Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.476Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/api/utils.distributed.html</loc>
-    <lastmod>2025-08-04T00:09:05.727Z</lastmod>
+    <lastmod>2025-08-04T14:27:14.979Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/docs/multipack.html</loc>
-    <lastmod>2025-08-04T00:05:24.914Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.775Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/examples/colab-notebooks/colab-axolotl-example.html</loc>
-    <lastmod>2025-08-04T00:05:24.918Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.779Z</lastmod>
   </url>
   <url>
     <loc>https://docs.axolotl.ai/FAQS.html</loc>
-    <lastmod>2025-08-04T00:05:24.908Z</lastmod>
+    <lastmod>2025-08-04T14:23:58.769Z</lastmod>
   </url>
 </urlset>