From 96b9c27cdff0c3c946b22a1f6156c2cff7136995 Mon Sep 17 00:00:00 2001 From: Svetlana Karslioglu Date: Thu, 22 Aug 2024 14:26:41 -0700 Subject: [PATCH 01/10] Add programmable Google Search to pytorch tutorials site (#2820) * Add programmable Google Search to pytorch tutorials site --- _static/css/custom.css | 9 +++++++++ _templates/layout.html | 17 +++++++++++++++++ 2 files changed, 26 insertions(+) diff --git a/_static/css/custom.css b/_static/css/custom.css index a467a08815..09aba28f25 100755 --- a/_static/css/custom.css +++ b/_static/css/custom.css @@ -91,3 +91,12 @@ transition: none; transform-origin: none; } + +.pytorch-left-menu-search input[type=text] { + background-image: none; +} + +.gsc-control-cse { + padding-left: 0px !important; + padding-bottom: 0px !important; +} diff --git a/_templates/layout.html b/_templates/layout.html index 22129040e4..1c632de63f 100644 --- a/_templates/layout.html +++ b/_templates/layout.html @@ -11,6 +11,23 @@ {%- endblock %} +{% block sidebartitle %} + {% if theme_display_version %} + {%- set nav_version = version %} + {% if READTHEDOCS and current_version %} + {%- set nav_version = current_version %} + {% endif %} + {% if nav_version %} +
+ {{ nav_version }} +
+ {% endif %} + {% endif %} +
+ +
+
+{% endblock %} {% block footer %} {{ super() }} From ea2dfc67b42c24d999edff0235702a314aab466a Mon Sep 17 00:00:00 2001 From: Ankith Gunapal Date: Fri, 23 Aug 2024 10:58:18 -0700 Subject: [PATCH 02/10] Tutorial for AOTI Python runtime (#2997) * Tutorial for AOTI Python runtime --------- Co-authored-by: Svetlana Karslioglu Co-authored-by: Angela Yi --- .ci/docker/build.sh | 3 +- .ci/docker/common/common_utils.sh | 2 +- .ci/docker/requirements.txt | 6 +- .jenkins/metadata.json | 3 + en-wordlist.txt | 3 +- recipes_source/recipes_index.rst | 6 + recipes_source/torch_export_aoti_python.py | 220 +++++++++++++++++++++ 7 files changed, 237 insertions(+), 6 deletions(-) create mode 100644 recipes_source/torch_export_aoti_python.py diff --git a/.ci/docker/build.sh b/.ci/docker/build.sh index 31f42fdbd8..c646b8f9a8 100755 --- a/.ci/docker/build.sh +++ b/.ci/docker/build.sh @@ -11,8 +11,9 @@ IMAGE_NAME="$1" shift export UBUNTU_VERSION="20.04" +export CUDA_VERSION="12.4.1" -export BASE_IMAGE="ubuntu:${UBUNTU_VERSION}" +export BASE_IMAGE="nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}" echo "Building ${IMAGE_NAME} Docker image" docker build \ diff --git a/.ci/docker/common/common_utils.sh b/.ci/docker/common/common_utils.sh index b20286a409..c7eabda555 100644 --- a/.ci/docker/common/common_utils.sh +++ b/.ci/docker/common/common_utils.sh @@ -22,5 +22,5 @@ conda_run() { } pip_install() { - as_ci_user conda run -n py_$ANACONDA_PYTHON_VERSION pip install --progress-bar off $* + as_ci_user conda run -n py_$ANACONDA_PYTHON_VERSION pip3 install --progress-bar off $* } diff --git a/.ci/docker/requirements.txt b/.ci/docker/requirements.txt index 00cf2f2103..9668b17fc3 100644 --- a/.ci/docker/requirements.txt +++ b/.ci/docker/requirements.txt @@ -30,8 +30,8 @@ pytorch-lightning torchx torchrl==0.5.0 tensordict==0.5.0 -ax-platform>==0.4.0 -nbformat>==5.9.2 +ax-platform>=0.4.0 +nbformat>=5.9.2 datasets transformers torchmultimodal-nightly # needs to be updated to stable as soon as it's avaialable @@ -68,4 +68,4 @@ pygame==2.1.2 pycocotools semilearn==0.3.2 torchao==0.0.3 -segment_anything==1.0 \ No newline at end of file +segment_anything==1.0 diff --git a/.jenkins/metadata.json b/.jenkins/metadata.json index 4814f9a7d2..2f1a9933aa 100644 --- a/.jenkins/metadata.json +++ b/.jenkins/metadata.json @@ -28,6 +28,9 @@ "intermediate_source/model_parallel_tutorial.py": { "needs": "linux.16xlarge.nvidia.gpu" }, + "recipes_source/torch_export_aoti_python.py": { + "needs": "linux.g5.4xlarge.nvidia.gpu" + }, "advanced_source/pendulum.py": { "needs": "linux.g5.4xlarge.nvidia.gpu", "_comment": "need to be here for the compiling_optimizer_lr_scheduler.py to run." diff --git a/en-wordlist.txt b/en-wordlist.txt index 62762ab69c..e69cbaa1a5 100644 --- a/en-wordlist.txt +++ b/en-wordlist.txt @@ -2,6 +2,7 @@ ACL ADI AOT +AOTInductor APIs ATen AVX @@ -617,4 +618,4 @@ warmstarting warmup webp wsi -wsis \ No newline at end of file +wsis diff --git a/recipes_source/recipes_index.rst b/recipes_source/recipes_index.rst index d94d7d5c22..caccdcc28f 100644 --- a/recipes_source/recipes_index.rst +++ b/recipes_source/recipes_index.rst @@ -150,6 +150,12 @@ Recipes are bite-sized, actionable examples of how to use specific PyTorch featu :link: ../recipes/recipes/swap_tensors.html :tags: Basics +.. customcarditem:: + :header: torch.export AOTInductor Tutorial for Python runtime + :card_description: Learn an end-to-end example of how to use AOTInductor for python runtime. + :image: ../_static/img/thumbnails/cropped/generic-pytorch-logo.png + :link: ../recipes/torch_export_aoti_python.html + :tags: Basics .. Interpretability diff --git a/recipes_source/torch_export_aoti_python.py b/recipes_source/torch_export_aoti_python.py new file mode 100644 index 0000000000..136862078c --- /dev/null +++ b/recipes_source/torch_export_aoti_python.py @@ -0,0 +1,220 @@ +# -*- coding: utf-8 -*- + +""" +(Beta) ``torch.export`` AOTInductor Tutorial for Python runtime +=============================================================== +**Author:** Ankith Gunapal, Bin Bao, Angela Yi +""" + +###################################################################### +# +# .. warning:: +# +# ``torch._inductor.aot_compile`` and ``torch._export.aot_load`` are in Beta status and are subject to backwards compatibility +# breaking changes. This tutorial provides an example of how to use these APIs for model deployment using Python runtime. +# +# It has been shown `previously `__ how AOTInductor can be used +# to do Ahead-of-Time compilation of PyTorch exported models by creating +# a shared library that can be run in a non-Python environment. +# +# +# In this tutorial, you will learn an end-to-end example of how to use AOTInductor for python runtime. +# We will look at how to use :func:`torch._inductor.aot_compile` along with :func:`torch.export.export` to generate a +# shared library. Additionally, we will examine how to execute the shared library in Python runtime using :func:`torch._export.aot_load`. +# You will learn about the speed up seen in the first inference time using AOTInductor, especially when using +# ``max-autotune`` mode which can take some time to execute. +# +# **Contents** +# +# .. contents:: +# :local: + +###################################################################### +# Prerequisites +# ------------- +# * PyTorch 2.4 or later +# * Basic understanding of ``torch.export`` and AOTInductor +# * Complete the `AOTInductor: Ahead-Of-Time Compilation for Torch.Export-ed Models `_ tutorial + +###################################################################### +# What you will learn +# ---------------------- +# * How to use AOTInductor for python runtime. +# * How to use :func:`torch._inductor.aot_compile` along with :func:`torch.export.export` to generate a shared library +# * How to run a shared library in Python runtime using :func:`torch._export.aot_load`. +# * When do you use AOTInductor for python runtime + +###################################################################### +# Model Compilation +# ----------------- +# +# We will use the TorchVision pretrained `ResNet18` model and TorchInductor on the +# exported PyTorch program using :func:`torch._inductor.aot_compile`. +# +# .. note:: +# +# This API also supports :func:`torch.compile` options like ``mode`` +# This means that if used on a CUDA enabled device, you can, for example, set ``"max_autotune": True`` +# which leverages Triton based matrix multiplications & convolutions, and enables CUDA graphs by default. +# +# We also specify ``dynamic_shapes`` for the batch dimension. In this example, ``min=2`` is not a bug and is +# explained in `The 0/1 Specialization Problem `__ + + +import os +import torch +from torchvision.models import ResNet18_Weights, resnet18 + +model = resnet18(weights=ResNet18_Weights.DEFAULT) +model.eval() + +with torch.inference_mode(): + + # Specify the generated shared library path + aot_compile_options = { + "aot_inductor.output_path": os.path.join(os.getcwd(), "resnet18_pt2.so"), + } + if torch.cuda.is_available(): + device = "cuda" + aot_compile_options.update({"max_autotune": True}) + else: + device = "cpu" + + model = model.to(device=device) + example_inputs = (torch.randn(2, 3, 224, 224, device=device),) + + # min=2 is not a bug and is explained in the 0/1 Specialization Problem + batch_dim = torch.export.Dim("batch", min=2, max=32) + exported_program = torch.export.export( + model, + example_inputs, + # Specify the first dimension of the input x as dynamic + dynamic_shapes={"x": {0: batch_dim}}, + ) + so_path = torch._inductor.aot_compile( + exported_program.module(), + example_inputs, + # Specify the generated shared library path + options=aot_compile_options + ) + + +###################################################################### +# Model Inference in Python +# ------------------------- +# +# Typically, the shared object generated above is used in a non-Python environment. In PyTorch 2.3, +# we added a new API called :func:`torch._export.aot_load` to load the shared library in the Python runtime. +# The API follows a structure similar to the :func:`torch.jit.load` API . You need to specify the path +# of the shared library and the device where it should be loaded. +# +# .. note:: +# In the example above, we specified ``batch_size=1`` for inference and it still functions correctly even though we specified ``min=2`` in +# :func:`torch.export.export`. + + +import os +import torch + +device = "cuda" if torch.cuda.is_available() else "cpu" +model_so_path = os.path.join(os.getcwd(), "resnet18_pt2.so") + +model = torch._export.aot_load(model_so_path, device) +example_inputs = (torch.randn(1, 3, 224, 224, device=device),) + +with torch.inference_mode(): + output = model(example_inputs) + +###################################################################### +# When to use AOTInductor for Python Runtime +# ------------------------------------------ +# +# One of the requirements for using AOTInductor is that the model shouldn't have any graph breaks. +# Once this requirement is met, the primary use case for using AOTInductor Python Runtime is for +# model deployment using Python. +# There are mainly two reasons why you would use AOTInductor Python Runtime: +# +# - ``torch._inductor.aot_compile`` generates a shared library. This is useful for model +# versioning for deployments and tracking model performance over time. +# - With :func:`torch.compile` being a JIT compiler, there is a warmup +# cost associated with the first compilation. Your deployment needs to account for the +# compilation time taken for the first inference. With AOTInductor, the compilation is +# done offline using ``torch.export.export`` & ``torch._indutor.aot_compile``. The deployment +# would only load the shared library using ``torch._export.aot_load`` and run inference. +# +# +# The section below shows the speedup achieved with AOTInductor for first inference +# +# We define a utility function ``timed`` to measure the time taken for inference +# + +import time +def timed(fn): + # Returns the result of running `fn()` and the time it took for `fn()` to run, + # in seconds. We use CUDA events and synchronization for accurate + # measurement on CUDA enabled devices. + if torch.cuda.is_available(): + start = torch.cuda.Event(enable_timing=True) + end = torch.cuda.Event(enable_timing=True) + start.record() + else: + start = time.time() + + result = fn() + if torch.cuda.is_available(): + end.record() + torch.cuda.synchronize() + else: + end = time.time() + + # Measure time taken to execute the function in miliseconds + if torch.cuda.is_available(): + duration = start.elapsed_time(end) + else: + duration = (end - start) * 1000 + + return result, duration + + +###################################################################### +# Lets measure the time for first inference using AOTInductor + +torch._dynamo.reset() + +model = torch._export.aot_load(model_so_path, device) +example_inputs = (torch.randn(1, 3, 224, 224, device=device),) + +with torch.inference_mode(): + _, time_taken = timed(lambda: model(example_inputs)) + print(f"Time taken for first inference for AOTInductor is {time_taken:.2f} ms") + + +###################################################################### +# Lets measure the time for first inference using ``torch.compile`` + +torch._dynamo.reset() + +model = resnet18(weights=ResNet18_Weights.DEFAULT).to(device) +model.eval() + +model = torch.compile(model) +example_inputs = torch.randn(1, 3, 224, 224, device=device) + +with torch.inference_mode(): + _, time_taken = timed(lambda: model(example_inputs)) + print(f"Time taken for first inference for torch.compile is {time_taken:.2f} ms") + +###################################################################### +# We see that there is a drastic speedup in first inference time using AOTInductor compared +# to ``torch.compile`` + +###################################################################### +# Conclusion +# ---------- +# +# In this recipe, we have learned how to effectively use the AOTInductor for Python runtime by +# compiling and loading a pretrained ``ResNet18`` model using the ``torch._inductor.aot_compile`` +# and ``torch._export.aot_load`` APIs. This process demonstrates the practical application of +# generating a shared library and running it within a Python environment, even with dynamic shape +# considerations and device-specific optimizations. We also looked at the advantage of using +# AOTInductor in model deployments, with regards to speed up in first inference time. From ec867b2f0a0764fb5139277247d8d07c76a82fe4 Mon Sep 17 00:00:00 2001 From: Svetlana Karslioglu Date: Sat, 24 Aug 2024 11:14:07 -0700 Subject: [PATCH 03/10] Create tutorial_submission_policy.md (#2995) - Add a policy for new tutorials submissions --------- Co-authored-by: albanD Co-authored-by: Nikita Shulga <2453524+malfet@users.noreply.github.com> --- README.md | 2 + tutorial_submission_policy.md | 107 ++++++++++++++++++++++++++++++++++ 2 files changed, 109 insertions(+) create mode 100644 tutorial_submission_policy.md diff --git a/README.md b/README.md index 0c961afd26..fe4b4b6edd 100644 --- a/README.md +++ b/README.md @@ -22,6 +22,8 @@ We use sphinx-gallery's [notebook styled examples](https://sphinx-gallery.github Here is how you can create a new tutorial (for a detailed description, see [CONTRIBUTING.md](./CONTRIBUTING.md)): +NOTE: Before submitting a new tutorial, read [PyTorch Tutorial Submission Policy](./tutorial_submission_policy.md). + 1. Create a Python file. If you want it executed while inserted into documentation, save the file with the suffix `tutorial` so that the file name is `your_tutorial.py`. 2. Put it in one of the `beginner_source`, `intermediate_source`, `advanced_source` directory based on the level of difficulty. If it is a recipe, add it to `recipes_source`. For tutorials demonstrating unstable prototype features, add to the `prototype_source`. 3. For Tutorials (except if it is a prototype feature), include it in the `toctree` directive and create a `customcarditem` in [index.rst](./index.rst). diff --git a/tutorial_submission_policy.md b/tutorial_submission_policy.md new file mode 100644 index 0000000000..c5c3a80087 --- /dev/null +++ b/tutorial_submission_policy.md @@ -0,0 +1,107 @@ +# PyTorch Tutorial Submission Policy + +This policy outlines the criteria and process for submitting new +tutorials to the PyTorch community. +Our goal is to ensure that all tutorials are of high quality, +relevant, and up-to-date, supporting both the growth of the PyTorch +users and the evolution of the PyTorch framework itself. By following +these guidelines, contributors can help us maintain a robust and +informative educational environment. + +## Acceptance Criteria For New Tutorials + +We accept new tutorials that adhere to one of the following use cases: + +* **Demonstrate New PyTorch Features:** Tutorials that support new features + for upcoming PyTorch releases are typically authored by the engineers who + are developing these features. These tutorials are crucial for showcasing + the latest advancements in PyTorch. We typically do not require more than + one tutorial per feature. + +* **Tutorials showcasing PyTorch usage with other tools and libraries:** We + accept community-contributed tutorials that illustrate innovative uses of + PyTorch alongside other open-source projects, models, and tools. Please + ensure that your tutorial remains neutral and does not promote or endorse + proprietary technologies over others. + +The first use case does not require going through the submission +process outlined below. If your tutorial falls under the second category, +please read and follow the instructions in the +**Submission Process For Community-Contributed Tutorials** section. + +## Submission Process For Community-Contributed Tutorials + +To maintain the quality and relevance of tutorials, we request that +community-contributed tutorials undergo a review process. If you are +interested in contributing a tutorial, please follow these steps: + +1. **Create an issue:** + * Open an issue in the pytorch/tutorials repository proposing the + new tutorial. Clearly explain the importance of the tutorial and + confirm that there is no existing tutorial covering the same or + similar topic. A tutorial should not disproportionately endorse + one technology over another. Please consult with Core Maintainers + to ensure your content adheres to these guidelines. + Use the provided [ISSUE_TEMPLATE](/~https://github.com/pytorch/tutorials/blob/main/.github/ISSUE_TEMPLATE/feature-request.yml) for the new tutorial request - select **Feature request** when submitting an issue. + + * If there is an existing tutorial on the topic that you would + like to significantly refactor, you can submit a PR. In the + description of the PR, explain why the changes are needed and + how they improve the tutorial. + + * These issues will be triaged by PyTorch maintainers on a case-by-case basis. + * Link any supporting materials including discussions in other repositories. + +1. **Await Approval:** + * Wait for a response from the PyTorch Tutorials maintainers. A PyTorch + tutorial maintainer will review your proposal and + determine whether a tutorial on the proposed topic is desirable. + A comment and an **approved** label will be added to your issue + by a maintainer. The review process for new tutorial PRs submitted + without the corresponding issue may take longer. + +1. **Adhere to writing and styling guidelines:** + * Once approved, follow the guidelines outlined in [CONTRIBUTING.md](/~https://github.com/pytorch/tutorials/blob/main/CONTRIBUTING.md) + and use the provided [template](/~https://github.com/pytorch/tutorials/blob/main/beginner_source/template_tutorial.py) for creating your tutorial. + * Link the issue in which you received approval for your tutorial + in the PR. + * We accept tutorials in both ``.rst`` (ReStructuredText) and ``.py`` + (Python) formats. However, unless your tutorial involves using + multiple GPU, parallel/distributed training, or requires extended + execution time (25 minutes or more), we prefer submissions + in Python file format. + +## Maintaining Tutorials + +When you submit a new tutorial, we encourage you to keep it in sync +with the latest PyTorch updates and features. Additionally, we may +contact you to review any PRs, issues, and other related matters to +ensure the tutorial remains a valuable resource. + +Please note the following: + +* If a tutorial breaks against the main branch, it will + be excluded from the build and an issue will be filed against it, + with the author/maintainer notified. If the issue is not resolved + within 90 days, the tutorial might be deleted from the repository. + +* We recommend that each tutorial is reviewed at least once a year to + ensure its relevance. + +## Deleting Stale Tutorials + +A tutorial might be considered stale when it no longer aligns with +the latest PyTorch updates, features, or best practices or best +practices: + +* The tutorial is no longer functional due to changes in PyTorch or + its dependencies +* The tutorial has been superseded by a newer, more comprehensive, or + more accurate tutorial +* The tutorial does not run successfully in the (CI), indicating + potential compatibility or dependency issues. + +If a tutorial is deemed stale, we will attempt to contact the code owner, +or someone from the tutorial mainatainers might attempt to update it. +However, if despite those attempts we fail to fix it, the tutorial +might be removed from the repository. From 4c1dd9d3fa1e6b41ef53dfc49ae5338fd6ce28cd Mon Sep 17 00:00:00 2001 From: Tim Statler Date: Mon, 26 Aug 2024 19:02:14 -0700 Subject: [PATCH 04/10] Removed upper-case letter/made 'download' the link text instead of 'here'/identified zip file (#3015) --- beginner_source/chatbot_tutorial.py | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/beginner_source/chatbot_tutorial.py b/beginner_source/chatbot_tutorial.py index 44310cc362..f902f8cd71 100644 --- a/beginner_source/chatbot_tutorial.py +++ b/beginner_source/chatbot_tutorial.py @@ -84,8 +84,7 @@ # Preparations # ------------ # -# To start, Download the data ZIP file -# `here `__ +# To get started, `download `__ the Movie-Dialogs Corpus zip file. # and put in a ``data/`` directory under the current directory. # From afcd669d56554c2ef46f2da92edbfaf8c54043ec Mon Sep 17 00:00:00 2001 From: Svetlana Karslioglu Date: Tue, 27 Aug 2024 12:57:19 -0700 Subject: [PATCH 05/10] Add weights_only=True to torch.load (#3012) * Add weights_only=True to torch.load --- advanced_source/dynamic_quantization_tutorial.py | 3 ++- advanced_source/static_quantization_tutorial.rst | 2 +- beginner_source/basics/quickstart_tutorial.py | 2 +- beginner_source/basics/saveloadrun_tutorial.py | 16 +++++++++++++--- beginner_source/blitz/cifar10_tutorial.py | 2 +- beginner_source/fgsm_tutorial.py | 2 +- beginner_source/saving_loading_models.py | 16 ++++++++-------- beginner_source/transfer_learning_tutorial.py | 2 +- .../autograd_saved_tensors_hooks_tutorial.py | 6 +++--- intermediate_source/ddp_tutorial.rst | 2 +- intermediate_source/tiatoolbox_tutorial.rst | 2 +- prototype_source/fx_graph_mode_ptq_dynamic.py | 3 ++- prototype_source/fx_graph_mode_ptq_static.rst | 6 +++--- prototype_source/pt2e_quant_ptq.rst | 2 +- prototype_source/pt2e_quant_qat.rst | 2 +- .../intel_neural_compressor_for_pytorch.rst | 2 +- .../recipes/module_load_state_dict_tips.py | 8 ++++---- .../recipes/save_load_across_devices.py | 2 +- .../saving_and_loading_a_general_checkpoint.py | 2 +- .../saving_and_loading_models_for_inference.py | 2 +- .../saving_multiple_models_in_one_file.py | 2 +- ...el_using_parameters_from_a_different_model.py | 2 +- 22 files changed, 50 insertions(+), 38 deletions(-) diff --git a/advanced_source/dynamic_quantization_tutorial.py b/advanced_source/dynamic_quantization_tutorial.py index 9cc07a1d95..c8d94789d5 100644 --- a/advanced_source/dynamic_quantization_tutorial.py +++ b/advanced_source/dynamic_quantization_tutorial.py @@ -151,7 +151,8 @@ def tokenize(self, path): model.load_state_dict( torch.load( model_data_filepath + 'word_language_model_quantize.pth', - map_location=torch.device('cpu') + map_location=torch.device('cpu'), + weights_only=True ) ) diff --git a/advanced_source/static_quantization_tutorial.rst b/advanced_source/static_quantization_tutorial.rst index 3b818aa03a..efb171c0df 100644 --- a/advanced_source/static_quantization_tutorial.rst +++ b/advanced_source/static_quantization_tutorial.rst @@ -286,7 +286,7 @@ We next define several helper functions to help with model evaluation. These mos def load_model(model_file): model = MobileNetV2() - state_dict = torch.load(model_file) + state_dict = torch.load(model_file, weights_only=True) model.load_state_dict(state_dict) model.to('cpu') return model diff --git a/beginner_source/basics/quickstart_tutorial.py b/beginner_source/basics/quickstart_tutorial.py index 07a1be517d..df7628081b 100644 --- a/beginner_source/basics/quickstart_tutorial.py +++ b/beginner_source/basics/quickstart_tutorial.py @@ -216,7 +216,7 @@ def test(dataloader, model, loss_fn): # the state dictionary into it. model = NeuralNetwork().to(device) -model.load_state_dict(torch.load("model.pth")) +model.load_state_dict(torch.load("model.pth", weights_only=True)) ############################################################# # This model can now be used to make predictions. diff --git a/beginner_source/basics/saveloadrun_tutorial.py b/beginner_source/basics/saveloadrun_tutorial.py index 16a9f03741..5b3aef124b 100644 --- a/beginner_source/basics/saveloadrun_tutorial.py +++ b/beginner_source/basics/saveloadrun_tutorial.py @@ -32,9 +32,14 @@ ########################## # To load model weights, you need to create an instance of the same model first, and then load the parameters # using ``load_state_dict()`` method. +# +# In the code below, we set ``weights_only=True`` to limit the +# functions executed during unpickling to only those necessary for +# loading weights. Using ``weights_only=True`` is considered +# a best practice when loading weights. model = models.vgg16() # we do not specify ``weights``, i.e. create untrained model -model.load_state_dict(torch.load('model_weights.pth')) +model.load_state_dict(torch.load('model_weights.pth', weights_only=True)) model.eval() ########################### @@ -50,9 +55,14 @@ torch.save(model, 'model.pth') ######################## -# We can then load the model like this: +# We can then load the model as demonstrated below. +# +# As described in `Saving and loading torch.nn.Modules `__, +# saving ``state_dict``s is considered the best practice. However, +# below we use ``weights_only=False`` because this involves loading the +# model, which is a legacy use case for ``torch.save``. -model = torch.load('model.pth') +model = torch.load('model.pth', weights_only=False), ######################## # .. note:: This approach uses Python `pickle `_ module when serializing the model, thus it relies on the actual class definition to be available when loading the model. diff --git a/beginner_source/blitz/cifar10_tutorial.py b/beginner_source/blitz/cifar10_tutorial.py index 8e3f325292..f38abdd566 100644 --- a/beginner_source/blitz/cifar10_tutorial.py +++ b/beginner_source/blitz/cifar10_tutorial.py @@ -221,7 +221,7 @@ def forward(self, x): # wasn't necessary here, we only did it to illustrate how to do so): net = Net() -net.load_state_dict(torch.load(PATH)) +net.load_state_dict(torch.load(PATH, weights_only=True)) ######################################################################## # Okay, now let us see what the neural network thinks these examples above are: diff --git a/beginner_source/fgsm_tutorial.py b/beginner_source/fgsm_tutorial.py index 007ad3fd95..9bdf52d84b 100644 --- a/beginner_source/fgsm_tutorial.py +++ b/beginner_source/fgsm_tutorial.py @@ -192,7 +192,7 @@ def forward(self, x): model = Net().to(device) # Load the pretrained model -model.load_state_dict(torch.load(pretrained_model, map_location=device)) +model.load_state_dict(torch.load(pretrained_model, map_location=device, weights_only=True)) # Set the model in evaluation mode. In this case this is for the Dropout layers model.eval() diff --git a/beginner_source/saving_loading_models.py b/beginner_source/saving_loading_models.py index fcd33be253..6c9b6b1fd7 100644 --- a/beginner_source/saving_loading_models.py +++ b/beginner_source/saving_loading_models.py @@ -153,7 +153,7 @@ # .. code:: python # # model = TheModelClass(*args, **kwargs) -# model.load_state_dict(torch.load(PATH)) +# model.load_state_dict(torch.load(PATH), weights_only=True) # model.eval() # # .. note:: @@ -206,7 +206,7 @@ # .. code:: python # # # Model class must be defined somewhere -# model = torch.load(PATH) +# model = torch.load(PATH, weights_only=False) # model.eval() # # This save/load process uses the most intuitive syntax and involves the @@ -290,7 +290,7 @@ # model = TheModelClass(*args, **kwargs) # optimizer = TheOptimizerClass(*args, **kwargs) # -# checkpoint = torch.load(PATH) +# checkpoint = torch.load(PATH, weights_only=True) # model.load_state_dict(checkpoint['model_state_dict']) # optimizer.load_state_dict(checkpoint['optimizer_state_dict']) # epoch = checkpoint['epoch'] @@ -354,7 +354,7 @@ # optimizerA = TheOptimizerAClass(*args, **kwargs) # optimizerB = TheOptimizerBClass(*args, **kwargs) # -# checkpoint = torch.load(PATH) +# checkpoint = torch.load(PATH, weights_only=True) # modelA.load_state_dict(checkpoint['modelA_state_dict']) # modelB.load_state_dict(checkpoint['modelB_state_dict']) # optimizerA.load_state_dict(checkpoint['optimizerA_state_dict']) @@ -407,7 +407,7 @@ # .. code:: python # # modelB = TheModelBClass(*args, **kwargs) -# modelB.load_state_dict(torch.load(PATH), strict=False) +# modelB.load_state_dict(torch.load(PATH), strict=False, weights_only=True) # # Partially loading a model or loading a partial model are common # scenarios when transfer learning or training a new complex model. @@ -446,7 +446,7 @@ # # device = torch.device('cpu') # model = TheModelClass(*args, **kwargs) -# model.load_state_dict(torch.load(PATH, map_location=device)) +# model.load_state_dict(torch.load(PATH, map_location=device, weights_only=True)) # # When loading a model on a CPU that was trained with a GPU, pass # ``torch.device('cpu')`` to the ``map_location`` argument in the @@ -469,7 +469,7 @@ # # device = torch.device("cuda") # model = TheModelClass(*args, **kwargs) -# model.load_state_dict(torch.load(PATH)) +# model.load_state_dict(torch.load(PATH, weights_only=True)) # model.to(device) # # Make sure to call input = input.to(device) on any input tensors that you feed to the model # @@ -497,7 +497,7 @@ # # device = torch.device("cuda") # model = TheModelClass(*args, **kwargs) -# model.load_state_dict(torch.load(PATH, map_location="cuda:0")) # Choose whatever GPU device number you want +# model.load_state_dict(torch.load(PATH, weights_only=True, map_location="cuda:0")) # Choose whatever GPU device number you want # model.to(device) # # Make sure to call input = input.to(device) on any input tensors that you feed to the model # diff --git a/beginner_source/transfer_learning_tutorial.py b/beginner_source/transfer_learning_tutorial.py index 7a2b053763..de7a178bd7 100644 --- a/beginner_source/transfer_learning_tutorial.py +++ b/beginner_source/transfer_learning_tutorial.py @@ -209,7 +209,7 @@ def train_model(model, criterion, optimizer, scheduler, num_epochs=25): print(f'Best val Acc: {best_acc:4f}') # load best model weights - model.load_state_dict(torch.load(best_model_params_path)) + model.load_state_dict(torch.load(best_model_params_path, weights_only=True)) return model diff --git a/intermediate_source/autograd_saved_tensors_hooks_tutorial.py b/intermediate_source/autograd_saved_tensors_hooks_tutorial.py index f16b170ee6..ed581426c2 100644 --- a/intermediate_source/autograd_saved_tensors_hooks_tutorial.py +++ b/intermediate_source/autograd_saved_tensors_hooks_tutorial.py @@ -397,7 +397,7 @@ def pack_hook(tensor): return name def unpack_hook(name): - return torch.load(name) + return torch.load(name, weights_only=True) ###################################################################### @@ -420,7 +420,7 @@ def pack_hook(tensor): return name def unpack_hook(name): - tensor = torch.load(name) + tensor = torch.load(name, weights_only=True) os.remove(name) return tensor @@ -462,7 +462,7 @@ def pack_hook(tensor): return temp_file def unpack_hook(temp_file): - return torch.load(temp_file.name) + return torch.load(temp_file.name, weights_only=True) ###################################################################### diff --git a/intermediate_source/ddp_tutorial.rst b/intermediate_source/ddp_tutorial.rst index 13297fb2a1..cff5105fa5 100644 --- a/intermediate_source/ddp_tutorial.rst +++ b/intermediate_source/ddp_tutorial.rst @@ -214,7 +214,7 @@ and elasticity support, please refer to `TorchElastic `_. diff --git a/recipes_source/recipes/module_load_state_dict_tips.py b/recipes_source/recipes/module_load_state_dict_tips.py index 17c812b016..70e9830cb3 100644 --- a/recipes_source/recipes/module_load_state_dict_tips.py +++ b/recipes_source/recipes/module_load_state_dict_tips.py @@ -39,7 +39,7 @@ def forward(self, x): # to ``torch.load``, the ``torch.device()`` context manager and the ``assign`` # keyword argument to ``nn.Module.load_state_dict()``. -state_dict = torch.load('checkpoint.pth', mmap=True) +state_dict = torch.load('checkpoint.pth', mmap=True, weights_only=True) with torch.device('meta'): meta_m = SomeModule(1000) meta_m.load_state_dict(state_dict, assign=True) @@ -47,7 +47,7 @@ def forward(self, x): ############################################################################# # Compare the snippet below to the one above: -state_dict = torch.load('checkpoint.pth') +state_dict = torch.load('checkpoint.pth', weights_only=True) m = SomeModule(1000) m.load_state_dict(state_dict) @@ -71,7 +71,7 @@ def forward(self, x): # * Waiting for the entire checkpoint to be loaded into RAM before performing, for example, some per-tensor processing. start_time = time.time() -state_dict = torch.load('checkpoint.pth') +state_dict = torch.load('checkpoint.pth', weights_only=True) end_time = time.time() print(f"loading time without mmap={end_time - start_time}") @@ -84,7 +84,7 @@ def forward(self, x): # storages will be memory-mapped. start_time = time.time() -state_dict = torch.load('checkpoint.pth', mmap=True) +state_dict = torch.load('checkpoint.pth', mmap=True, weights_only=True) end_time = time.time() print(f"loading time with mmap={end_time - start_time}") diff --git a/recipes_source/recipes/save_load_across_devices.py b/recipes_source/recipes/save_load_across_devices.py index be950e15b1..c59af8821e 100644 --- a/recipes_source/recipes/save_load_across_devices.py +++ b/recipes_source/recipes/save_load_across_devices.py @@ -97,7 +97,7 @@ def forward(self, x): # Load device = torch.device('cpu') model = Net() -model.load_state_dict(torch.load(PATH, map_location=device)) +model.load_state_dict(torch.load(PATH, map_location=device, weights_only=True)) ###################################################################### diff --git a/recipes_source/recipes/saving_and_loading_a_general_checkpoint.py b/recipes_source/recipes/saving_and_loading_a_general_checkpoint.py index 31b14f3a28..8c773a1490 100644 --- a/recipes_source/recipes/saving_and_loading_a_general_checkpoint.py +++ b/recipes_source/recipes/saving_and_loading_a_general_checkpoint.py @@ -131,7 +131,7 @@ def forward(self, x): model = Net() optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9) -checkpoint = torch.load(PATH) +checkpoint = torch.load(PATH, weights_only=True) model.load_state_dict(checkpoint['model_state_dict']) optimizer.load_state_dict(checkpoint['optimizer_state_dict']) epoch = checkpoint['epoch'] diff --git a/recipes_source/recipes/saving_and_loading_models_for_inference.py b/recipes_source/recipes/saving_and_loading_models_for_inference.py index cd24b77c1d..7adce2a90b 100644 --- a/recipes_source/recipes/saving_and_loading_models_for_inference.py +++ b/recipes_source/recipes/saving_and_loading_models_for_inference.py @@ -117,7 +117,7 @@ def forward(self, x): # Load model = Net() -model.load_state_dict(torch.load(PATH)) +model.load_state_dict(torch.load(PATH, weights_only=True)) model.eval() diff --git a/recipes_source/recipes/saving_multiple_models_in_one_file.py b/recipes_source/recipes/saving_multiple_models_in_one_file.py index f468d7ac6a..e938be03b4 100644 --- a/recipes_source/recipes/saving_multiple_models_in_one_file.py +++ b/recipes_source/recipes/saving_multiple_models_in_one_file.py @@ -128,7 +128,7 @@ def forward(self, x): optimModelA = optim.SGD(modelA.parameters(), lr=0.001, momentum=0.9) optimModelB = optim.SGD(modelB.parameters(), lr=0.001, momentum=0.9) -checkpoint = torch.load(PATH) +checkpoint = torch.load(PATH, weights_only=True) modelA.load_state_dict(checkpoint['modelA_state_dict']) modelB.load_state_dict(checkpoint['modelB_state_dict']) optimizerA.load_state_dict(checkpoint['optimizerA_state_dict']) diff --git a/recipes_source/recipes/warmstarting_model_using_parameters_from_a_different_model.py b/recipes_source/recipes/warmstarting_model_using_parameters_from_a_different_model.py index 40aeeea9db..a0752bfc67 100644 --- a/recipes_source/recipes/warmstarting_model_using_parameters_from_a_different_model.py +++ b/recipes_source/recipes/warmstarting_model_using_parameters_from_a_different_model.py @@ -124,7 +124,7 @@ def forward(self, x): # are loading into. # -netB.load_state_dict(torch.load(PATH), strict=False) +netB.load_state_dict(torch.load(PATH, weights_only=True), strict=False) ###################################################################### From fc016bdc4b10cdbe3437114279dcdcb73fadd82f Mon Sep 17 00:00:00 2001 From: dev_thomas <36235705+hadh93@users.noreply.github.com> Date: Thu, 29 Aug 2024 01:41:22 +0900 Subject: [PATCH 06/10] Fix typos in dynamic_quantization_bert_tutorial.rst (#3019) --- intermediate_source/dynamic_quantization_bert_tutorial.rst | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/intermediate_source/dynamic_quantization_bert_tutorial.rst b/intermediate_source/dynamic_quantization_bert_tutorial.rst index 1ea6ea46dd..e515f53a1d 100644 --- a/intermediate_source/dynamic_quantization_bert_tutorial.rst +++ b/intermediate_source/dynamic_quantization_bert_tutorial.rst @@ -79,7 +79,7 @@ Mac: .. code:: shell - yes y | pip uninstall torch tochvision + yes y | pip uninstall torch torchvision yes y | pip install --pre torch -f https://download.pytorch.org/whl/nightly/cu101/torch_nightly.html @@ -206,7 +206,7 @@ in `examples `_. +We provide the fine-tuned BERT model for MRPC task `here `_. To save time, you can download the model file (~400 MB) directly into your local folder ``$OUT_DIR``. 2.1 Set global configurations @@ -273,7 +273,7 @@ We load the tokenizer and fine-tuned BERT sequence classifier model 2.3 Define the tokenize and evaluation function ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ -We reuse the tokenize and evaluation function from `Huggingface `_. +We reuse the tokenize and evaluation function from `HuggingFace `_. .. code:: python From 4cb73cabeeae16d52d35e3098312da123ed5b431 Mon Sep 17 00:00:00 2001 From: Richard Zou Date: Thu, 29 Aug 2024 10:05:07 -0400 Subject: [PATCH 07/10] Improve custom ops tutorials (#3020) Co-authored-by: Svetlana Karslioglu --- advanced_source/cpp_custom_ops.rst | 2 ++ advanced_source/python_custom_ops.py | 15 ++++++++++----- 2 files changed, 12 insertions(+), 5 deletions(-) diff --git a/advanced_source/cpp_custom_ops.rst b/advanced_source/cpp_custom_ops.rst index 435ff088bc..ffabd6eff7 100644 --- a/advanced_source/cpp_custom_ops.rst +++ b/advanced_source/cpp_custom_ops.rst @@ -174,6 +174,8 @@ To add ``torch.compile`` support for an operator, we must add a FakeTensor kerne known as a "meta kernel" or "abstract impl"). FakeTensors are Tensors that have metadata (such as shape, dtype, device) but no data: the FakeTensor kernel for an operator specifies how to compute the metadata of output tensors given the metadata of input tensors. +The FakeTensor kernel should return dummy Tensors of your choice with +the correct Tensor metadata (shape/strides/``dtype``/device). We recommend that this be done from Python via the `torch.library.register_fake` API, though it is possible to do this from C++ as well (see diff --git a/advanced_source/python_custom_ops.py b/advanced_source/python_custom_ops.py index 1e429b76b3..0b3bf6e474 100644 --- a/advanced_source/python_custom_ops.py +++ b/advanced_source/python_custom_ops.py @@ -66,7 +66,7 @@ def display(img): ###################################################################### # ``crop`` is not handled effectively out-of-the-box by # ``torch.compile``: ``torch.compile`` induces a -# `"graph break" `_ +# `"graph break" `_ # on functions it is unable to handle and graph breaks are bad for performance. # The following code demonstrates this by raising an error # (``torch.compile`` with ``fullgraph=True`` raises an error if a @@ -85,9 +85,9 @@ def f(img): # # 1. wrap the function into a PyTorch custom operator. # 2. add a "``FakeTensor`` kernel" (aka "meta kernel") to the operator. -# Given the metadata (e.g. shapes) -# of the input Tensors, this function says how to compute the metadata -# of the output Tensor(s). +# Given some ``FakeTensors`` inputs (dummy Tensors that don't have storage), +# this function should return dummy Tensors of your choice with the correct +# Tensor metadata (shape/strides/``dtype``/device). from typing import Sequence @@ -130,6 +130,11 @@ def f(img): # ``autograd.Function`` with PyTorch operator registration APIs can lead to (and # has led to) silent incorrectness when composed with ``torch.compile``. # +# If you don't need training support, there is no need to use +# ``torch.library.register_autograd``. +# If you end up training with a ``custom_op`` that doesn't have an autograd +# registration, we'll raise an error message. +# # The gradient formula for ``crop`` is essentially ``PIL.paste`` (we'll leave the # derivation as an exercise to the reader). Let's first wrap ``paste`` into a # custom operator: @@ -203,7 +208,7 @@ def setup_context(ctx, inputs, output): ###################################################################### # Mutable Python Custom operators # ------------------------------- -# You can also wrap a Python function that mutates its inputs into a custom +# You can also wrap a Python function that mutates its inputs into a custom # operator. # Functions that mutate inputs are common because that is how many low-level # kernels are written; for example, a kernel that computes ``sin`` may take in From 2bdd6420081f40c3d467adb147ee22fb74c6e828 Mon Sep 17 00:00:00 2001 From: Tim Statler Date: Thu, 29 Aug 2024 08:13:14 -0700 Subject: [PATCH 08/10] Removed outdated steps in README about running about setup.py (#3014) Co-authored-by: Svetlana Karslioglu --- README.md | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/README.md b/README.md index fe4b4b6edd..af84d9ebe7 100644 --- a/README.md +++ b/README.md @@ -33,7 +33,7 @@ If you are starting off with a Jupyter notebook, you can use [this script](https ## Building locally -The tutorial build is very large and requires a GPU. If your machine does not have a GPU device, you can preview your HTML build without actually downloading the data and running the tutorial code: +The tutorial build is very large and requires a GPU. If your machine does not have a GPU device, you can preview your HTML build without actually downloading the data and running the tutorial code: 1. Install required dependencies by running: `pip install -r requirements.txt`. @@ -42,8 +42,6 @@ The tutorial build is very large and requires a GPU. If your machine does not ha - If you have a GPU-powered laptop, you can build using `make docs`. This will download the data, execute the tutorials and build the documentation to `docs/` directory. This might take about 60-120 min for systems with GPUs. If you do not have a GPU installed on your system, then see next step. - You can skip the computationally intensive graph generation by running `make html-noplot` to build basic html documentation to `_build/html`. This way, you can quickly preview your tutorial. -> If you get **ModuleNotFoundError: No module named 'pytorch_sphinx_theme' make: *** [html-noplot] Error 2** from /tutorials/src/pytorch-sphinx-theme or /venv/src/pytorch-sphinx-theme (while using virtualenv), run `python setup.py install`. - ## Building a single tutorial You can build a single tutorial by using the `GALLERY_PATTERN` environment variable. For example to run only `neural_style_transfer_tutorial.py`, run: @@ -61,8 +59,8 @@ The `GALLERY_PATTERN` variable respects regular expressions. ## About contributing to PyTorch Documentation and Tutorials -* You can find information about contributing to PyTorch documentation in the -PyTorch Repo [README.md](/~https://github.com/pytorch/pytorch/blob/master/README.md) file. +* You can find information about contributing to PyTorch documentation in the +PyTorch Repo [README.md](/~https://github.com/pytorch/pytorch/blob/master/README.md) file. * Additional information can be found in [PyTorch CONTRIBUTING.md](/~https://github.com/pytorch/pytorch/blob/master/CONTRIBUTING.md). From b3c49a3f5e42e22db8982a9ecb3e1d5b30b83ec0 Mon Sep 17 00:00:00 2001 From: Svetlana Karslioglu Date: Thu, 29 Aug 2024 09:56:59 -0700 Subject: [PATCH 09/10] Fix hovering over the GCS search button (#3005) * Fix hovering over the GCS search button --- _static/css/custom.css | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/_static/css/custom.css b/_static/css/custom.css index 09aba28f25..cc195d9906 100755 --- a/_static/css/custom.css +++ b/_static/css/custom.css @@ -100,3 +100,15 @@ padding-left: 0px !important; padding-bottom: 0px !important; } + +.gsc-search-button .gsc-search-button-v2:focus { + border: transparent !important; + outline: none; + box-shadow: none; +} +.gsc-search-button-v2:active { + border: none !important; +} +.gsc-search-button-v2 { + border: none !important; +} From e07d43bb7b9b3e84c5682eb5b0cb8c07e2d35051 Mon Sep 17 00:00:00 2001 From: Tim Statler Date: Fri, 30 Aug 2024 14:36:59 -0700 Subject: [PATCH 10/10] Added warnings to select Pytorch mobile tutorials directing users to ExecuTorch (#3016) * Added warnings to select ExecuTorch tutorials/recipes/prototypes * Added warnings to select ExecuTorch tutorials/recipes/prototypes * Added redirect for renamed prototype * Update deeplabv3_on_android.rst Fixed misplaced info tag. * Apply suggestions from code review --------- Co-authored-by: Svetlana Karslioglu --- beginner_source/deeplabv3_on_android.rst | 4 ++++ prototype_source/lite_interpreter.rst | 9 +++++++++ recipes_source/mobile_interpreter.rst | 3 +++ recipes_source/mobile_perf.rst | 5 ++++- recipes_source/ptmobile_recipes_summary.rst | 3 +++ 5 files changed, 23 insertions(+), 1 deletion(-) create mode 100644 prototype_source/lite_interpreter.rst diff --git a/beginner_source/deeplabv3_on_android.rst b/beginner_source/deeplabv3_on_android.rst index f2fe0e48f1..5ca7f01ad0 100644 --- a/beginner_source/deeplabv3_on_android.rst +++ b/beginner_source/deeplabv3_on_android.rst @@ -5,6 +5,10 @@ Image Segmentation DeepLabV3 on Android **Reviewed by**: `Jeremiah Chung `_ +.. warning:: + PyTorch Mobile is no longer actively supported. Please check out `ExecuTorch `_, PyTorch’s all-new on-device inference library. You can also review our `end-to-end workflows `_ and review the `source code for DeepLabV3 `_. + + Introduction ------------ diff --git a/prototype_source/lite_interpreter.rst b/prototype_source/lite_interpreter.rst new file mode 100644 index 0000000000..73e950d72e --- /dev/null +++ b/prototype_source/lite_interpreter.rst @@ -0,0 +1,9 @@ +(Prototype) Introduce lite interpreter workflow in Android and iOS +======================= + +This tutorial has been moved to https://pytorch.org/tutorials/recipes/mobile_interpreter.html + + +.. raw:: html + + diff --git a/recipes_source/mobile_interpreter.rst b/recipes_source/mobile_interpreter.rst index dda1dd9243..44036e74ff 100644 --- a/recipes_source/mobile_interpreter.rst +++ b/recipes_source/mobile_interpreter.rst @@ -3,6 +3,9 @@ **Author**: `Chen Lai `_, `Martin Yuan `_ +.. warning:: + PyTorch Mobile is no longer actively supported. Please check out `ExecuTorch `_, PyTorch’s all-new on-device inference library. You can also review our new documentation to learn more about how to build `iOS `_ and `Android `_ apps with ExecuTorch. + Introduction ------------ diff --git a/recipes_source/mobile_perf.rst b/recipes_source/mobile_perf.rst index aae1447cbf..14f183ab69 100644 --- a/recipes_source/mobile_perf.rst +++ b/recipes_source/mobile_perf.rst @@ -1,6 +1,9 @@ Pytorch Mobile Performance Recipes ================================== +.. warning:: + PyTorch Mobile is no longer actively supported. Please check out `ExecuTorch `_, PyTorch’s all-new on-device inference library. You can also learn more about `quantization `_, `Hardware acceleration (op fusion using hw) `_, and `benchmarking `_ on ExecuTorch’s documentation pages. + Introduction ---------------- Performance (aka latency) is crucial to most, if not all, @@ -245,7 +248,7 @@ For example, using ResNet-50 and running the following script: -you would get the following result: +you would get the following result: :: diff --git a/recipes_source/ptmobile_recipes_summary.rst b/recipes_source/ptmobile_recipes_summary.rst index cddee940f2..6cc8f6f751 100644 --- a/recipes_source/ptmobile_recipes_summary.rst +++ b/recipes_source/ptmobile_recipes_summary.rst @@ -1,6 +1,9 @@ Summary of PyTorch Mobile Recipes ===================================== +.. warning:: + Note: PyTorch Mobile is no longer actively supported. Please check out `ExecuTorch `_, PyTorch’s all-new on-device inference library. You can also review these `ExecuTorch examples `_. + This summary provides a top level overview of recipes for PyTorch Mobile to help developers choose which recipes to follow for their PyTorch-powered mobile app development. Introduction