Merge branch 'main' into jafraust/ddp

Author: jafraustro

.ci/docker/requirements.txt

@@ -23,7 +23,7 @@ tqdm==4.66.1
 numpy==1.24.4
 matplotlib
 librosa
-torch==2.7
+torch==2.8
 torchvision
 torchdata
 networkx
@@ -37,8 +37,8 @@ tensorboard
 jinja2==3.1.3
 pytorch-lightning
 torchx
-torchrl==0.7.2
-tensordict==0.7.2
+torchrl==0.9.2
+tensordict==0.9.1
 # For ax_multiobjective_nas_tutorial.py
 ax-platform>=0.4.0,<0.5.0
 nbformat>=5.9.2

.jenkins/validate_tutorials_built.py

@@ -20,6 +20,7 @@
     "beginner_source/examples_nn/polynomial_optim",
     "beginner_source/examples_autograd/polynomial_autograd",
     "beginner_source/examples_autograd/polynomial_custom_function",
+    "intermediate_source/dqn_with_rnn_tutorial", #not working on 2.8 release reenable after 3514
     "intermediate_source/mnist_train_nas",  # used by ax_multiobjective_nas_tutorial.py
     "intermediate_source/torch_compile_conv_bn_fuser",
     "intermediate_source/_torch_export_nightly_tutorial",  # does not work on release

_static/img/thumbnails/cropped/visualizing_gradients_tutorial.png

@@ -60,7 +60,7 @@ were not merged or issues that have been closed without a merged PR.
 | kiszk | 20 | https://github.com/pytorch/pytorch/pull/128337, https://github.com/pytorch/pytorch/pull/128123, https://github.com/pytorch/pytorch/pull/128022, https://github.com/pytorch/pytorch/pull/128312 |
 | loganthomas | 19 | https://github.com/pytorch/pytorch/pull/128676, https://github.com/pytorch/pytorch/pull/128192, https://github.com/pytorch/pytorch/pull/128189, https://github.com/pytorch/tutorials/pull/2922, https://github.com/pytorch/tutorials/pull/2910, https://github.com/pytorch/xla/pull/7195 |
 | ignaciobartol | 17 | https://github.com/pytorch/pytorch/pull/128741, https://github.com/pytorch/pytorch/pull/128135, https://github.com/pytorch/pytorch/pull/127938, https://github.com/pytorch/tutorials/pull/2936 |
-| arunppsg | 17 | https://github.com/pytorch/pytorch/pull/128391, https://github.com/pytorch/pytorch/pull/128021, https://github.com/pytorch/pytorch/pull/128018, https://github.com/pytorch-labs/torchfix/pull/59 |
+| arunppsg | 17 | https://github.com/pytorch/pytorch/pull/128391, https://github.com/pytorch/pytorch/pull/128021, https://github.com/pytorch/pytorch/pull/128018, https://github.com/meta-pytorch/torchfix/pull/59 |
 | alperenunlu | 17 | https://github.com/pytorch/tutorials/pull/2934, https://github.com/pytorch/tutorials/pull/2909, https://github.com/pytorch/pytorch/pull/104043 |
 | anandptl84 | 10 | https://github.com/pytorch/pytorch/pull/128196, https://github.com/pytorch/pytorch/pull/128098 |
 | GdoongMathew | 10 | https://github.com/pytorch/pytorch/pull/128136, https://github.com/pytorch/pytorch/pull/128051 |

docathon-leaderboard.md

@@ -3,10 +3,10 @@ Welcome to PyTorch Tutorials
 
 **What's new in PyTorch tutorials?**
 
-* `Utilizing Torch Function modes with torch.compile <https://pytorch.org/tutorials/recipes/torch_compile_torch_function_modes.html>`__
-* `Context Parallel Tutorial <https://pytorch.org/tutorials/prototype/context_parallel.html>`__
-* `(beta) Explicit horizontal fusion with foreach_map and torch.compile <https://pytorch.org/tutorials/recipes/foreach_map.html>`__
-* Updated `Inductor Windows CPU Tutorial <https://pytorch.org/tutorials/prototype/inductor_windows.html>`__
+* `Integrating Custom Operators with SYCL for Intel GPU <https://pytorch.org/tutorials/advanced/cpp_custom_ops_sycl.html>`__
+* `Supporting Custom C++ Classes in torch.compile/torch.export <https://docs.pytorch.org/tutorials/advanced/custom_class_pt2.html>`__
+* `Accelerating torch.save and torch.load with GPUDirect Storage <https://docs.pytorch.org/tutorials/unstable/gpu_direct_storage.html>`__
+* `Getting Started with Fully Sharded Data Parallel (FSDP2) <https://docs.pytorch.org/tutorials/intermediate/FSDP_tutorial.html>`__
 
 .. raw:: html
 
@@ -24,7 +24,7 @@ Welcome to PyTorch Tutorials
 .. customcalloutitem::
    :description: Bite-size, ready-to-deploy PyTorch code examples.
    :header: PyTorch Recipes
-   :button_link: recipes/recipes_index.html
+   :button_link: recipes_index.html
    :button_text: Explore Recipes
 
 .. End of callout item section
@@ -99,6 +99,13 @@ Welcome to PyTorch Tutorials
    :link: intermediate/pinmem_nonblock.html
    :tags: Getting-Started
 
+.. customcarditem::
+   :header: Visualizing Gradients in PyTorch
+   :card_description: Visualize the gradient flow of a network.
+   :image: _static/img/thumbnails/cropped/visualizing_gradients_tutorial.png
+   :link: intermediate/visualizing_gradients_tutorial.html
+   :tags: Getting-Started
+
 .. Image/Video
 
 .. customcarditem::
@@ -700,14 +707,14 @@ Welcome to PyTorch Tutorials
    :header: Building an ExecuTorch iOS Demo App
    :card_description: Explore how to set up the ExecuTorch iOS Demo App, which uses the MobileNet v3 model to process live camera images leveraging three different backends: XNNPACK, Core ML, and Metal Performance Shaders (MPS).
    :image: _static/img/ExecuTorch-Logo-cropped.svg
-   :link: https://github.com/pytorch-labs/executorch-examples/tree/main/mv3/apple/ExecuTorchDemo
+   :link: https://github.com/meta-pytorch/executorch-examples/tree/main/mv3/apple/ExecuTorchDemo
    :tags: Edge
 
 .. customcarditem::
    :header: Building an ExecuTorch Android Demo App
    :card_description: Learn how to set up the ExecuTorch Android Demo App for image segmentation tasks using the DeepLab v3 model and XNNPACK FP32 backend.
    :image: _static/img/ExecuTorch-Logo-cropped.svg
-   :link: https://github.com/pytorch-labs/executorch-examples/tree/main/dl3/android/DeepLabV3Demo#executorch-android-demo-app
+   :link: https://github.com/meta-pytorch/executorch-examples/tree/main/dl3/android/DeepLabV3Demo#executorch-android-demo-app
    :tags: Edge
 
 .. customcarditem::
@@ -837,4 +844,4 @@ Additional Resources
    :maxdepth: 1
    :hidden:
 
-   unstable_index
+   prototype/prototype_index

index.rst

@@ -62,7 +62,7 @@
 
 If you are only interested in performant attention score modifications, please
 check out the `FlexAttention blog <https://pytorch.org/blog/flexattention/>`_ that
-contains a `gym of masks <https://github.com/pytorch-labs/attention-gym>`_.
+contains a `gym of masks <https://github.com/meta-pytorch/attention-gym>`_.
 
 """
 
@@ -71,7 +71,7 @@
 # ===============================
 # First, we will briefly introduce the four technologies mentioned in the introduction
 #
-# * `torch.nested <https://pytorch.org/tutorials/prototype/nestedtensor.html>`_
+# * `torch.nested <https://pytorch.org/tutorials/unstable/nestedtensor.html>`_
 #
 # Nested tensors generalize the shape of regular dense tensors, allowing for
 # representation of ragged-sized data with the same tensor UX. In the context of
@@ -157,7 +157,7 @@
 #   skipped, performance and memory usage improve.
 #
 # We'll demonstrate the above by building upon the ``MultiheadAttention`` layer in the
-# `Nested Tensor tutorial <https://pytorch.org/tutorials/prototype/nestedtensor.html>`_
+# `Nested Tensor tutorial <https://pytorch.org/tutorials/unstable/nestedtensor.html>`_
 # and comparing it to the ``nn.MultiheadAttention`` layer.
 
 import torch
@@ -675,7 +675,7 @@ def benchmark(func, *args, **kwargs):
 # of the ``MultiheadAttention`` layer that allows for arbitrary modifications
 # to the attention score. The example below takes the ``alibi_mod``
 # that implements `ALiBi <https://arxiv.org/abs/2108.12409>`_ from
-# `attention gym <https://github.com/pytorch-labs/attention-gym>`_ and uses it
+# `attention gym <https://github.com/meta-pytorch/attention-gym>`_ and uses it
 # with nested input tensors.
 
 from torch.nn.attention.flex_attention import flex_attention
@@ -892,8 +892,8 @@ def forward(self, x):
 # etc. Further, there are several good examples of using various performant building blocks to
 # implement various transformer architectures. Some examples include
 #
-# * `gpt-fast <https://github.com/pytorch-labs/gpt-fast>`_
-# * `segment-anything-fast <https://github.com/pytorch-labs/segment-anything-fast>`_
+# * `gpt-fast <https://github.com/meta-pytorch/gpt-fast>`_
+# * `segment-anything-fast <https://github.com/meta-pytorch/segment-anything-fast>`_
 # * `lucidrains implementation of NaViT with nested tensors <https://github.com/lucidrains/vit-pytorch/blob/73199ab486e0fad9eced2e3350a11681db08b61b/vit_pytorch/na_vit_nested_tensor.py>`_
 # * `torchtune's implementation of VisionTransformer <https://github.com/pytorch/torchtune/blob/a8a64ec6a99a6ea2be4fdaf0cd5797b03a2567cf/torchtune/modules/vision_transformer.py#L16>`_