diff --git a/docs/2.2/_modules/torch.html b/docs/2.2/_modules/torch.html
index b55aa90e5219..c468ad5a48e8 100644
--- a/docs/2.2/_modules/torch.html
+++ b/docs/2.2/_modules/torch.html
@@ -2209,7 +2209,7 @@ <h1>Source code for torch</h1><div class="highlight"><pre>
 
 <span class="sd">        - Experimental or debug in-tree backends can be seen with `torch._dynamo.list_backends(None)`</span>
 
-<span class="sd">        - To register an out-of-tree custom backend: https://pytorch.org/docs/main/compile/custom-backends.html</span>
+<span class="sd">        - To register an out-of-tree custom backend: https://pytorch.org/docs/stable/torch.compiler_custom_backends.html</span>
 <span class="sd">       mode (str): Can be either &quot;default&quot;, &quot;reduce-overhead&quot;, &quot;max-autotune&quot; or &quot;max-autotune-no-cudagraphs&quot;</span>
 
 <span class="sd">        - &quot;default&quot; is the default mode, which is a good balance between performance and overhead</span>
diff --git a/docs/2.2/generated/torch.compile.html b/docs/2.2/generated/torch.compile.html
index f3c7ee227844..0b93e216cf26 100644
--- a/docs/2.2/generated/torch.compile.html
+++ b/docs/2.2/generated/torch.compile.html
@@ -498,7 +498,7 @@ <h1>torch.compile<a class="headerlink" href="#torch-compile" title="Permalink to
 <li><p>”inductor” is the default backend, which is a good balance between performance and overhead</p></li>
 <li><p>Non experimental in-tree backends can be seen with <cite>torch._dynamo.list_backends()</cite></p></li>
 <li><p>Experimental or debug in-tree backends can be seen with <cite>torch._dynamo.list_backends(None)</cite></p></li>
-<li><p>To register an out-of-tree custom backend: <a class="reference external" href="https://pytorch.org/docs/main/compile/custom-backends.html">https://pytorch.org/docs/main/compile/custom-backends.html</a></p></li>
+<li><p>To register an out-of-tree custom backend: <a class="reference external" href="https://pytorch.org/docs/stable/torch.compiler_custom_backends.html">https://pytorch.org/docs/stable/torch.compiler_custom_backends.html</a></p></li>
 </ul>
 </p></li>
 <li><p><strong>mode</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.12)"><em>str</em></a>) – <p>Can be either “default”, “reduce-overhead”, “max-autotune” or “max-autotune-no-cudagraphs”</p>
diff --git a/docs/2.3/_modules/torch.html b/docs/2.3/_modules/torch.html
index 72fcdadd16a6..87222b0c3479 100644
--- a/docs/2.3/_modules/torch.html
+++ b/docs/2.3/_modules/torch.html
@@ -2281,7 +2281,7 @@ <h1>Source code for torch</h1><div class="highlight"><pre>
 
 <span class="sd">        - Experimental or debug in-tree backends can be seen with `torch._dynamo.list_backends(None)`</span>
 
-<span class="sd">        - To register an out-of-tree custom backend: https://pytorch.org/docs/main/compile/custom-backends.html</span>
+<span class="sd">        - To register an out-of-tree custom backend: https://pytorch.org/docs/stable/torch.compiler_custom_backends.html</span>
 <span class="sd">       mode (str): Can be either &quot;default&quot;, &quot;reduce-overhead&quot;, &quot;max-autotune&quot; or &quot;max-autotune-no-cudagraphs&quot;</span>
 
 <span class="sd">        - &quot;default&quot; is the default mode, which is a good balance between performance and overhead</span>
diff --git a/docs/2.3/generated/torch.compile.html b/docs/2.3/generated/torch.compile.html
index c878a8d56416..8d7f038db67a 100644
--- a/docs/2.3/generated/torch.compile.html
+++ b/docs/2.3/generated/torch.compile.html
@@ -505,7 +505,7 @@ <h1>torch.compile<a class="headerlink" href="#torch-compile" title="Permalink to
 <li><p>”inductor” is the default backend, which is a good balance between performance and overhead</p></li>
 <li><p>Non experimental in-tree backends can be seen with <cite>torch._dynamo.list_backends()</cite></p></li>
 <li><p>Experimental or debug in-tree backends can be seen with <cite>torch._dynamo.list_backends(None)</cite></p></li>
-<li><p>To register an out-of-tree custom backend: <a class="reference external" href="https://pytorch.org/docs/main/compile/custom-backends.html">https://pytorch.org/docs/main/compile/custom-backends.html</a></p></li>
+<li><p>To register an out-of-tree custom backend: <a class="reference external" href="https://pytorch.org/docs/stable/torch.compiler_custom_backends.html">https://pytorch.org/docs/stable/torch.compiler_custom_backends.html</a></p></li>
 </ul>
 </p></li>
 <li><p><strong>mode</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.12)"><em>str</em></a>) – <p>Can be either “default”, “reduce-overhead”, “max-autotune” or “max-autotune-no-cudagraphs”</p>