[tutorial] typo fix, update torch.compiler_cudagraph_trees.md (#167713)

luyaor · pytorchmergebot · commit dbe61249eac1 · 2025-11-24T03:02:06.000Z
Fix a API typo in the cuda graph tutorial. The API given in cuda graph tutorial is wrong. Pull Request resolved: #167713 Approved by: https://github.com/jerryzh168
diff --git a/docs/source/torch.compiler_cudagraph_trees.md b/docs/source/torch.compiler_cudagraph_trees.md
@@ -319,7 +319,7 @@ Trees, we don’t want to add unintended dependencies between iterations that wo
 to prematurely free memory from a prior invocation. Our heuristics are in inference we start a new iteration on each invocation for
 torch.compile, and in training we do the same so long as there is not a pending backward that has not been invoked. If those heuristics
 are wrong, you can mark the start of a new iteration with
-[torch.compiler.mark_step_begin()](https://pytorch.org/docs/stable/generated/torch.compiler.cudagraph_mark_step_begin.html), or clone
+[torch.compiler.cudagraph_mark_step_begin()](https://pytorch.org/docs/stable/generated/torch.compiler.cudagraph_mark_step_begin.html), or clone
 tensors of a prior iteration (outside of torch.compile) before you begin the next run.
 
 ### Comparisons
