Merge branch 'main' into RC-TEST-2.9

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@f576ce406429f8ce11475ec096dda346df5b3b71#egg=pytorch_sphinx_theme2
+-e git+https://github.com/pytorch/pytorch_sphinx_theme.git@c2e38b37f3c432c610639f06d1d421c6df4c225c#egg=pytorch_sphinx_theme2
 
 # Tutorial dependencies
 tqdm==4.66.1

advanced_source/cpp_frontend.rst

@@ -1,11 +1,31 @@
+.. _cpp-frontend-tutorial:
+
 Using the PyTorch C++ Frontend
 ==============================
 
+**Author:** `Peter Goldsborough <https://github.com/goldsborough>`_
+
+.. grid:: 2
+
+    .. grid-item-card:: :octicon:`mortar-board;1em;` What you will learn
+       :class-card: card-prerequisites
+
+       * How to build a C++ application that utilizes the PyTorch C++ frontend
+       * How to define and train neural networks from C++ using PyTorch abstractions
+
+    .. grid-item-card:: :octicon:`list-unordered;1em;` Prerequisites
+       :class-card: card-prerequisites
+
+       * PyTorch 1.5 or later
+       * Basic understanding of C++ programming
+       * Basic Ubuntu Linux environment with CMake >= 3.5; similar commands will work in a MacOS / Windows environment
+       * (Optional) A CUDA-based GPU for the GPU training sections
+
 The PyTorch C++ frontend is a pure C++ interface to the PyTorch machine learning
 framework. While the primary interface to PyTorch naturally is Python, this
 Python API sits atop a substantial C++ codebase providing foundational data
 structures and functionality such as tensors and automatic differentiation. The
-C++ frontend exposes a pure C++11 API that extends this underlying C++ codebase
+C++ frontend exposes a pure C++17 API that extends this underlying C++ codebase
 with tools required for machine learning training and inference. This includes a
 built-in collection of common components for neural network modeling; an API to
 extend this collection with custom modules; a library of popular optimization
@@ -137,14 +157,14 @@ on we'll use this ``CMakeLists.txt`` file:
 
 .. code-block:: cmake
 
-  cmake_minimum_required(VERSION 3.0 FATAL_ERROR)
+  cmake_minimum_required(VERSION 3.5 FATAL_ERROR)
   project(dcgan)
 
   find_package(Torch REQUIRED)
 
   add_executable(dcgan dcgan.cpp)
   target_link_libraries(dcgan "${TORCH_LIBRARIES}")
-  set_property(TARGET dcgan PROPERTY CXX_STANDARD 14)
+  set_property(TARGET dcgan PROPERTY CXX_STANDARD 17)
 
 .. note::
 
@@ -859,7 +879,7 @@ stacks them into a single tensor along the first dimension:
 
 Note that the MNIST dataset should be located in the ``./mnist`` directory
 relative to wherever you execute the training binary from. You can use `this
-script <https://gist.github.com/goldsborough/6dd52a5e01ed73a642c1e772084bcd03>`_
+script <https://gist.github.com/jbschlosser/94347505df6188f8764793ee29fd1bdd>`_
 to download the MNIST dataset.
 
 Next, we create a data loader and pass it this dataset. To make a new data

beginner_source/introyt/tensors_deeper_tutorial.py

@@ -85,7 +85,7 @@
 
 
 #########################################################################
-# The fctory methods all do just what you’d expect - we have a tensor
+# The factory methods all do just what you’d expect - we have a tensor
 # full of zeros, another full of ones, and another with random values
 # between 0 and 1.
 # 

conf.py

@@ -145,7 +145,7 @@ def wrapper(*args, **kwargs):
 
 intersphinx_mapping = {
     "torch": ("https://docs.pytorch.org/docs/stable/", None),
-    "tensordict": ("https://docs.pytorch.github.io/tensordict/stable", None),
+    "tensordict": ("https://docs.pytorch.org/tensordict/stable", None),
     "torchrl": ("https://docs.pytorch.org/rl/stable", None),
     "torchaudio": ("https://docs.pytorch.org/audio/stable/", None),
     "torchtext": ("https://docs.pytorch.org/text/stable/", None),
@@ -236,6 +236,7 @@ def wrapper(*args, **kwargs):
     "navbar_center": "navbar-nav",
     "display_version": True,
     "pytorch_project": "tutorials",
+    "canonical_url": "https://docs.pytorch.org/tutorials/",
 }
 
 theme_variables = pytorch_sphinx_theme2.get_theme_variables()

intermediate_source/neural_tangent_kernels.py

@@ -13,7 +13,7 @@
 
 .. note::
 
-   This tutorial requires PyTorch 2.0.0 or later.
+   This tutorial requires PyTorch 2.6.0 or later.
 
 Setup
 -----
@@ -24,7 +24,12 @@
 import torch
 import torch.nn as nn
 from torch.func import functional_call, vmap, vjp, jvp, jacrev
-device = 'cuda' if torch.cuda.device_count() > 0 else 'cpu'
+
+if torch.accelerator.is_available() and torch.accelerator.device_count() > 0:
+    device = torch.accelerator.current_accelerator()
+else:
+    device = torch.device("cpu")
+
 
 class CNN(nn.Module):
     def __init__(self):

intermediate_source/reinforcement_ppo.py

@@ -25,12 +25,12 @@
 
 We will cover six crucial components of TorchRL:
 
-* `environments <https://pytorch.org/rl/reference/envs.html>`__
-* `transforms <https://pytorch.org/rl/reference/envs.html#transforms>`__
-* `models (policy and value function) <https://pytorch.org/rl/reference/modules.html>`__
-* `loss modules <https://pytorch.org/rl/reference/objectives.html>`__
-* `data collectors <https://pytorch.org/rl/reference/collectors.html>`__
-* `replay buffers <https://pytorch.org/rl/reference/data.html#replay-buffers>`__
+* `environments <https://docs.pytorch.org/rl/stable/reference/envs.html>`__
+* `transforms <https://docs.pytorch.org/rl/stable/reference/envs.html#transforms>`__
+* `models (policy and value function) <https://docs.pytorch.org/rl/stable/reference/modules.html>`__
+* `loss modules <https://docs.pytorch.org/rl/stable/reference/objectives.html>`__
+* `data collectors <https://docs.pytorch.org/rl/stable/reference/collectors.html>`__
+* `replay buffers <https://docs.pytorch.org/rl/stable/reference/data.html#replay-buffers>`__
 
 """
 
@@ -466,7 +466,7 @@
 # Data collector
 # --------------
 #
-# TorchRL provides a set of `DataCollector classes <https://pytorch.org/rl/reference/collectors.html>`__.
+# TorchRL provides a set of `DataCollector classes <https://docs.pytorch.org/rl/stable/reference/collectors.html>`__.
 # Briefly, these classes execute three operations: reset an environment,
 # compute an action given the latest observation, execute a step in the environment,
 # and repeat the last two steps until the environment signals a stop (or reaches

recipes_index.rst

@@ -367,7 +367,6 @@ from our full-length tutorials.
    recipes/recipes/timer_quick_start
    recipes/torch_compile_backend_ipex
    recipes/zero_redundancy_optimizer
-   recipes/cuda_rpc
    recipes/distributed_comm_debug_mode
    recipes/torch_export_challenges_solutions
    recipes/recipes/benchmark

recipes_source/cuda_rpc.rst

@@ -31,6 +31,7 @@
     "prototype/pt2e_quantizer.html": "https://docs.pytorch.org/ao/main/tutorials_source/pt2e_quantizer.html",
     "prototype/quantization_in_pytorch_2_0_export_tutorial.html": "../index.html",
     "prototype/torchscript_freezing.html": "../index.html",
+    "recipes_source/cuda_rpc.rst": "../index.html",
     "receipes/fuse.html": "../index.html",
     "receipes/quantization.html": "../index.html",
     "receipes/receipes/dynamic_quantization.html": "../index.html",