Merge branch 'main' into jafraust/ddp

Author: svekars

.ci/docker/requirements.txt

@@ -16,7 +16,7 @@ pandocfilters==1.5.1
 markdown==3.8.2
 
 # PyTorch Theme
--e git+https://github.com/pytorch/pytorch_sphinx_theme.git@67aa61448e33cae9a345e012be355f8030d3c4f5#egg=pytorch_sphinx_theme2
+-e git+https://github.com/pytorch/pytorch_sphinx_theme.git@3066b6f62b0da01c52570c928281bca68287933d#egg=pytorch_sphinx_theme2
 
 # Tutorial dependencies
 tqdm==4.66.1

recipes_source/regional_aot.py

@@ -3,7 +3,7 @@
 Reducing AoT cold start compilation time with regional compilation
 ============================================================================
 
-**Author:** `Sayak Paul <https://github.com/sayakpaul>`_, `Charles Bensimon <https://github.com/cbensimon>`_, `Angela Yi <https://github.com/angelayi>`_
+**Author:** `Sayak Paul <https://huggingface.co/sayakpaul>`_, `Charles Bensimon <https://huggingface.co/cbensimon>`_, `Angela Yi <https://github.com/angelayi>`_
 
 In the `regional compilation recipe <https://docs.pytorch.org/tutorials/recipes/regional_compilation.html>`__, we showed
 how to reduce cold start compilation times while retaining (almost) full compilation benefits. This was demonstrated for
@@ -236,4 +236,6 @@ def aot_compile_load_model(regional=False) -> torch.nn.Module:
 #
 # This recipe shows how to control the cold start time when compiling your
 # model ahead-of-time. This becomes effective when your model has repeated
-# blocks, which is typically seen in large generative models.
+# blocks, which is typically seen in large generative models. We used this
+# recipe on various models to speed up real-time performance. Learn more
+# `here <https://huggingface.co/blog/zerogpu-aoti>`__.