<picture>
<source media="(prefers-color-scheme: dark)" srcset="https://github.com/pytorch/pytorch/raw/main/docs/source/_static/img/pytorch-logo-light.png" />
<source media="(prefers-color-scheme: light)" srcset="https://github.com/pytorch/pytorch/raw/main/docs/source/_static/img/pytorch-logo-dark.png" />
<img src="https://github.com/pytorch/pytorch/raw/main/docs/source/_static/img/pytorch-logo-dark.png" alt="PyTorch logo" />
</picture>
--------------------------------------------------------------------------------
PyTorch is a Python package that provides two high-level features:
- Tensor computation (like NumPy) with strong GPU acceleration
- Deep neural networks built on a tape-based autograd system
You can reuse your favorite Python packages such as NumPy, SciPy, and Cython to extend PyTorch when needed.
Our trunk health (Continuous Integration signals) can be found at [hud.pytorch.org](https://hud.pytorch.org/ci/pytorch/pytorch/main).
<!-- toc -->
- [More About PyTorch](#more-about-pytorch)
- [A GPU-Ready Tensor Library](#a-gpu-ready-tensor-library)
- [Dynamic Neural Networks: Tape-Based Autograd](#dynamic-neural-networks-tape-based-autograd)
- [Python First](#python-first)
- [Imperative Experiences](#imperative-experiences)
- [Fast and Lean](#fast-and-lean)
- [Extensions Without Pain](#extensions-without-pain)
- [Installation](#installation)
- [Binaries](#binaries)
- [NVIDIA Jetson Platforms](#nvidia-jetson-platforms)
- [From Source](#from-source)
- [Prerequisites](#prerequisites)
- [NVIDIA CUDA Support](#nvidia-cuda-support)
- [AMD ROCm Support](#amd-rocm-support)
- [Intel GPU Support](#intel-gpu-support)
- [Get the PyTorch Source](#get-the-pytorch-source)
- [Install Dependencies](#install-dependencies)
- [Install PyTorch](#install-pytorch)
- [Adjust Build Options (Optional)](#adjust-build-options-optional)
- [Docker Image](#docker-image)
- [Using pre-built images](#using-pre-built-images)
- [Building the image yourself](#building-the-image-yourself)
- [Previous Versions](#previous-versions)
- [Getting Started](#getting-started)
- [Resources](#resources)
- [Communication](#communication)
- [Releases and Contributing](#releases-and-contributing)
- [The Team](#the-team)
- [License](#license)
<!-- tocstop -->
## More About PyTorch
[Learn the basics of PyTorch](https://pytorch.org/tutorials/beginner/basics/intro.html)
At a granular level, PyTorch is a library that consists of the following components:
| Component | Description |
| ---- | --- |
| [**torch**](https://pytorch.org/docs/stable/torch.html) | A Tensor library like NumPy, with strong GPU support |
| [**torch.autograd**](https://pytorch.org/docs/stable/autograd.html) | A tape-based automatic differentiation library that supports all differentiable Tensor operations in torch |
| [**torch.nn**](https://pytorch.org/docs/stable/nn.html) | A neural networks library deeply integrated with autograd designed for maximum flexibility |
| [**torch.multiprocessing**](https://pytorch.org/docs/stable/multiprocessing.html) | Python multiprocessing, but with magical memory sharing of torch Tensors across processes. Useful for data loading and Hogwild training |
| [**torch.utils**](https://pytorch.org/docs/stable/data.html) | DataLoader and other utility functions for convenience |
Usually, PyTorch is used either as:
- A replacement for NumPy to use the power of GPUs.
- A deep learning research platform that provides maximum flexibility and speed.
Elaborating Further:
### A GPU-Ready Tensor Library
If you use NumPy, then you have used Tensors (a.k.a. ndarray).
PyTorch provides Tensors that can live either on the CPU or the GPU and accelerates the
computation by a huge amount.
We provide a wide variety of tensor routines to accelerate and fit your scientific computation needs
such as slicing, indexing, mathematical operations, linear algebra, reductions.
And they are fast!
### Dynamic Neural Networks: Tape-Based Autograd
PyTorch has a unique way of building neural networks: using and replaying a tape recorder.
Most frameworks such as TensorFlow, Theano, Caffe, and CNTK have a static view of the world.
One has to build a neural network and reuse the same structure again and again.
Changing the way the network behaves means that one has to start from scratch.
With PyTorch, we use a technique called reverse-mode auto-differentiation, which allows you to
change the way your network behaves arbitrarily with zero lag or overhead. Our inspiration comes
from several research papers on this topic, as well as current and past work such as
[torch-autograd](https://github.com/twitter/torch-autograd),
[autograd](https://github.com/HIPS/autograd),
[Chainer](https://chainer.org), etc.
While this technique is not unique to PyTorch, it's one of the fastest implementations of it to date.
You get the best of speed and flexibility for your crazy research.
### Python First
PyTorch is not a Python binding into a monolithic C++ framework.
It is built to be deeply integrated into Python.
You can use it naturally like you would use [NumPy](https://www.numpy.org/) / [SciPy](https://www.scipy.org/) / [scikit-learn](https://scikit-learn.org) etc.
You can write your new neural network layers in Python itself, using your favorite libraries
and use packages such as [Cython](https://cython.org/) and [Numba](https://numba.pydata.org/).
Our goal is to not reinvent the wheel where appropriate.
### Imperative Experiences
PyTorch is designed to be intuitive, linear in thought, and easy to use.
When you execute a line of code, it gets executed. There isn't an asynchronous view of the world.
When you drop into a debugger or receive error messages and stack traces, understanding them is straightforward.
The stack trace points to exactly where your code was defined.
We hope you never spend hours debugging your code because of bad stack traces or asynchronous and opaque execution engines.
### Fast and Lean
PyTorch has minimal framework overhead. We integrate acceleration libraries
such as [Intel MKL](https://software.intel.com/mkl) and NVIDIA ([cuDNN](https://developer.nvidia.com/cudnn), [NCCL](https://developer.nvidia.com/nccl)) to maximize speed.
At the core, its CPU and GPU Tensor and neural network backends
are mature and have been tested for years.
Hence, PyTorch is quite fast — whether you run small or large neural networks.
The memory usage in PyTorch is extremely efficient compared to Torch or some of the alternatives.
We've written custom memory allocators for the GPU to make sure that
your deep learning models are maximally memory efficient.
This enables you to train bigger deep learning models than before.
### Extensions Without Pain
Writing new neural network modules, or interfacing with PyTorch's Tensor API, was designed to be straightforward
and with minimal abstractions.
You can write new neural network layers in Python using the torch API
[or your favorite NumPy-based libraries such as SciPy](https://pytorch.org/tutorials/advanced/numpy_extensions_tutorial.html).
If you want to write your layers in C/C++, we provide a convenient extension API that is efficient and with minimal boilerplate.
No wrapper code needs to be written. You can see [a tutorial here](https://pytorch.org/tutorials/advanced/cpp_extension.html) and [an example here](https://github.com/pytorch/extension-cpp).
## Installation
### Binaries
Commands to install binaries via Conda or pip wheels are on our website: [https://pytorch.org/get-started/locally/](https://pytorch.org/get-started/locally/)
#### NVIDIA Jetson Platforms
Python wheels for NVIDIA's Jetson Nano, Jetson TX1/TX2, Jetson Xavier NX/AGX, and Jetson AGX Orin are provided [here](https://forums.developer.nvidia.com/t/pytorch-for-jetson-version-1-10-now-available/72048) and the L4T container is published [here](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/l4t-pytorch)
They require JetPack 4.2 and above, and [@dusty-nv](https://github.com/dusty-nv) and [@ptrblck](https://github.com/ptrblck) are maintaining them.
### From Source
#### Prerequisites
If you are installing from source, you will need:
- Python 3.11 or later
- A compiler that fully supports C++20, such as clang or gcc (gcc 11.3.0 or newer is required, on Linux)
- Visual Studio or Visual Studio Build Tool (Windows only)
- At least 10 GB of free disk space
- 30-60 minutes for the initial build (subsequent rebuilds are much faster)
\* PyTorch CI uses Visual C++ BuildTools, which come with Visual Studio Enterprise,
Professional, or Community Editions. You can also install the build tools from
https://visualstudio.microsoft.com/visual-cpp-build-tools/. The build tools *do not*
come with Visual Studio Code by default.
An example of environment setup is shown below:
* Linux:
```bash
$ source <CONDA_INSTALL_DIR>/bin/activate
$ conda create -y -n <CONDA_NAME>
$ conda activate <CONDA_NAME>
```
* Windows:
```bash
$ source <CONDA_INSTALL_DIR>\Scripts\activate.bat
$ conda create -y -n <CONDA_NAME>
$ conda activate <CONDA_NAME>
$ call "C:\Program Files\Microsoft Visual Studio\<VERSION>\Community\VC\Auxiliary\Build\vcvarsall.bat" x64
```
A conda environment is not required. You can also do a PyTorch build in a
standard virtual environment, e.g., created with tools like `uv`, provided
your system has installed all the necessary dependencies unavailable as pip
packages (e.g., CUDA, MKL.)
##### NVIDIA CUDA Support
If you want to compile with CUDA support, [select a supported version of CUDA from our support matrix](https://pytorch.org/get-started/locally/), then install the following:
- [NVIDIA CUDA](https://developer.nvidia.com/cuda-downloads)
- [NVIDIA cuDNN](https://developer.nvidia.com/cudnn) v9.0 or above
- [Compiler](https://gist.github.com/ax3l/9489132) compatible with CUDA
Note: You could refer to the [cuDNN Support Matrix](https://docs.nvidia.com/deeplearning/cudnn/backend/latest/reference/support-matrix.html) for cuDNN versions with the various supported CUDA, CUDA driver, and NVIDIA hardware.
If you want to disable CUDA support, export the environment variable `USE_CUDA=0`.
Other potentially useful environment variables are documented in [`cmake/EnvVarForwarding.cmake`](./cmake/EnvVarForwarding.cmake). If
CUDA is installed in a non-standard location, set PATH so that the nvcc you
want to use can be found (e.g., `export PATH=/usr/local/cuda-12.8/bin:$PATH`).
If you are building for NVIDIA's Jetson platforms (Jetson Nano, TX1, TX2, AGX Xavier), Instructions to install PyTorch for Jetson Nano are [available here](https://devtalk.nvidia.com/default/topic/1049071/jetson-nano/pytorch-for-jetson-nano/)
##### AMD ROCm Support
If you want to compile with ROCm support, install
- [AMD ROCm](https://rocm.docs.amd.com/en/latest/deploy/linux/quick_start.html) 4.0 and above installation
- ROCm is currently supported only for Linux systems.
By default the build system expects ROCm to be installed in `/opt/rocm`. If ROCm is installed in a different directory, the `ROCM_PATH` environment variable must be set to the ROCm installation directory. The build system automatically detects the AMD GPU architecture. Optionally, the AMD GPU architecture can be explicitly set with the `PYTORCH_ROCM_ARCH` environment variable [AMD GPU architecture](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html#supported-gpus)
If you want to disable ROCm support, export the environment variable `USE_ROCM=0`.
Other potentially useful environment variables are documented in [`cmake/EnvVarForwarding.cmake`](./cmake/EnvVarForwarding.cmake).
##### Intel GPU Support
If you want to compile with Intel GPU support, follow these
- [PyTorch Prerequisites for Intel GPUs](https://www.intel.com/content/www/us/en/developer/articles/tool/pytorch-prerequisites-for-intel-gpu.html) instructions.
- Intel GPU is supported for Linux and Windows.
If you want to disable Intel GPU support, export the environment variable `USE_XPU=0`.
Other potentially useful environment variables are documented in [`cmake/EnvVarForwarding.cmake`](./cmake/EnvVarForwarding.cmake).
#### Get the PyTorch Source
```bash
git clone https://github.com/pytorch/pytorch
cd pytorch
# if you are updating an existing checkout
git submodule sync
git submodule update --init --recursive
```
#### Install Dependencies
**Common**
```bash
# Run this command from the PyTorch directory after cloning the source code using the “Get the PyTorch Source“ section above
pip install --group dev
```
**On Linux**
```bash
pip install mkl-static mkl-include
# CUDA only: Add LAPACK support for the GPU if needed
# magma installation: run with active conda environment. specify CUDA version to install
.ci/docker/common/install_magma_conda.sh 12.4
# (optional) If using torch.compile with inductor/triton, install the matching version of triton
# Run from the pytorch directory after cloning
# For Intel GPU support, please explicitly `export USE_XPU=1` before running command.
make triton
```
**On Windows**
```bash
pip install mkl-static mkl-include
# Add these packages if torch.distributed is needed.
# Distributed package support on Windows is a prototype feature and is subject to changes.
conda install -c conda-forge libuv=1.51
```
#### Install PyTorch
**On Linux**
If you're compiling for AMD ROCm then first run this command:
```bash
# Only run this if you're compiling for ROCm
python tools/amd_build/build_amd.py
```
Install PyTorch
```bash
# the CMake prefix for conda environment
export CMAKE_PREFIX_PATH="${CONDA_PREFIX:-'$(dirname $(which conda))/../'}:${CMAKE_PREFIX_PATH}"
python -m pip install --no-build-isolation -v -e .
# the CMake prefix for non-conda environment, e.g. Python venv
# call following after activating the venv
export CMAKE_PREFIX_PATH="${VIRTUAL_ENV}:${CMAKE_PREFIX_PATH}"
```
**On macOS**
```bash
python -m pip install --no-build-isolation -v -e .
```
**On Windows**
If you want to build legacy python code, please refer to [Building on legacy code and CUDA](https://github.com/pytorch/pytorch/blob/main/CONTRIBUTING.md#building-on-legacy-code-and-cuda)
**CPU-only builds**
In this mode PyTorch computations will run on your CPU, not your GPU.
```cmd
python -m pip install --no-build-isolation -v -e .
```
Note on OpenMP: The desired OpenMP implementation is Intel OpenMP (iomp). In order to link against iomp, you'll need to manually download the library and set up the building environment by tweaking `CMAKE_INCLUDE_PATH` and `LIB`. The instruction [here](https://docs.pytorch.org/docs/main/notes/windows.html#building-from-source) is an example for setting up both MKL and Intel OpenMP. Without these configurations for CMake, Microsoft Visual C OpenMP runtime (vcomp) will be used.
**CUDA based build**
In this mode PyTorch computations will leverage your GPU via CUDA for faster number crunching
[NVTX](https://docs.nvidia.com/gameworks/content/gameworkslibrary/nvtx/nvidia_tools_extension_library_nvtx.htm) is needed to build PyTorch with CUDA.
NVTX is a part of CUDA distributive, where it is called "Nsight Compute". To install it onto an already installed CUDA run CUDA installation once again and check the corresponding checkbox.
Make sure that CUDA with Nsight Compute is installed after Visual Studio.
Currently, VS 2017 / 2019, and Ninja are supported as the generator of CMake. If `ninja.exe` is detected in `PATH`, then Ninja will be used as the default generator, otherwise, it will use VS 2017 / 2019.
<br/> If Ninja is selected as the generator, the latest MSVC will get selected as the underlying toolchain.
Additional libraries such as
[Magma](https://developer.nvidia.com/magma), [oneDNN, a.k.a. MKLDNN or DNNL](https://github.com/oneapi-src/oneDNN), and [Sccache](https://github.com/mozilla/sccache) are often needed. Please refer to the [installation-helper](https://github.com/pytorch/pytorch/tree/main/.ci/pytorch/win-test-helpers/installation-helpers) to install them.
You can refer to the [build_pytorch.bat](https://github.com/pytorch/pytorch/blob/main/.ci/pytorch/win-test-helpers/build_pytorch.bat) script for some other environment variables configurations
```cmd
cmd
:: Set the environment variables after you have downloaded and unzipped the mkl package,
:: else CMake would throw an error as `Could NOT find OpenMP`.
set CMAKE_INCLUDE_PATH={Your directory}\mkl\include
set LIB={Your directory}\mkl\lib;%LIB%
:: Read the content in the previous section carefully before you proceed.
:: [Optional] If you want to override the underlying toolset used by Ninja and Visual Studio with CUDA, please run the following script block.
:: "Visual Studio 2019 Developer Command Prompt" will be run automatically.
:: Make sure you have CMake >= 3.12 before you do this when you use the Visual Studio generator.
set CMAKE_GENERATOR_TOOLSET_VERSION=14.27
set DISTUTILS_USE_SDK=1
for /f "usebackq tokens=*" %i in (`"%ProgramFiles(x86)%\Microsoft Visual Studio\Installer\vswhere.exe" -version [15^,17^) -products * -latest -property installationPath`) do call "%i\VC\Auxiliary\Build\vcvarsall.bat" x64 -vcvars_ver=%CMAKE_GENERATOR_TOOLSET_VERSION%
:: [Optional] If you want to override the CUDA host compiler
set CUDAHOSTCXX=C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.27.29110\bin\HostX64\x64\cl.exe
python -m pip install --no-build-isolation -v -e .
```
**Intel GPU builds**
In this mode PyTorch with Intel GPU support will be built.
Please make sure [the common prerequisites](#prerequisites) as well as [the prerequisites for Intel GPU](#intel-gpu-support) are properly installed and the environment variables are configured prior to starting the build. For build tool support, `Visual Studio 2022` is required.
Then PyTorch can be built with the command:
```cmd
:: CMD Commands:
:: Set the CMAKE_PREFIX_PATH to help find corresponding packages
:: %CONDA_PREFIX% only works after `conda activate custom_env`
if defined CMAKE_PREFIX_PATH (
set "CMAKE_PREFIX_PATH=%CONDA_PREFIX%\Library;%CMAKE_PREFIX_PATH%"
) else (
set "CMAKE_PREFIX_PATH=%CONDA_PREFIX%\Library"
)
python -m pip install --no-build-isolation -v -e .
```
##### Adjust Build Options (Optional)
You can adjust the configuration of CMake variables through environment variables, which the
build forwards to CMake (see [`cmake/EnvVarForwarding.cmake`](./cmake/EnvVarForwarding.cmake) for
the full list). For example, pointing the build at a specific Conda prefix, CuDNN, or BLAS:
On Linux
```bash
export CMAKE_PREFIX_PATH="${CONDA_PREFIX:-'$(dirname $(which conda))/../'}:${CMAKE_PREFIX_PATH}"
spin develop
```
On macOS
```bash
export CMAKE_PREFIX_PATH="${CONDA_PREFIX:-'$(dirname $(which conda))/../'}:${CMAKE_PREFIX_PATH}"
MACOSX_DEPLOYMENT_TARGET=11.0 spin develop
```
After a first `spin develop`, you can inspect or adjust the CMake cache interactively with
`ccmake build` (or `cmake-gui build`) and rebuild with `spin develop`; your edits persist, because
reconfiguration does not use `--fresh` and environment variables only seed cache entries that are
not already set. The corollary: once a variable is in the cache, changing its environment variable
no longer affects it -- edit the cache directly (or delete `build/CMakeCache.txt`) to change it.
The build type and compiler are the exception; the build always re-applies them.
### Docker Image
#### Using pre-built images
You can also pull a pre-built docker image from Docker Hub and run with docker v23.0+
```bash
docker run --gpus all --rm -ti --ipc=host pytorch/pytorch:latest
```
Please note that PyTorch uses shared memory to share data between processes, so if torch multiprocessing is used (e.g.
for multithreaded data loaders) the default shared memory segment size that container runs with is not enough, and you
should increase shared memory size either with `--ipc=host` or `--shm-size` command line options to `nvidia-docker run`.
#### Building the image yourself
**NOTE:** Must be built with a Docker version >= 23.0
The Dockerfile is supplied to build images with CUDA 12.6 support and cuDNN v9.
You can pass `PYTHON_VERSION=x.y` make variable to specify which Python version is to be used by Miniconda, or leave it
unset to use the default, as the Dockerfile uses system Python.
```bash
make -f docker.Makefile
# images are tagged as docker.io/${your_docker_username}/pytorch
```
You can also pass the `CMAKE_VARS="..."` environment variable to specify additional CMake variables to be passed to CMake during the build.
See [`cmake/EnvVarForwarding.cmake`](./cmake/EnvVarForwarding.cmake) for the list of available variables.
```bash
make -f docker.Makefile
```
### Previous Versions
Installation instructions and binaries for previous PyTorch versions may be found
on [our website](https://pytorch.org/get-started/previous-versions).
## Getting Started
Pointers to get you started:
- [Tutorials: get you started with understanding and using PyTorch](https://pytorch.org/tutorials/)
- [Examples: easy to understand PyTorch code across all domains](https://github.com/pytorch/examples)
- [The API Reference](https://pytorch.org/docs/)
- [Glossary](https://github.com/pytorch/pytorch/blob/main/GLOSSARY.md)
## Resources
* [PyTorch.org](https://pytorch.org/)
* [PyTorch Tutorials](https://pytorch.org/tutorials/)
* [PyTorch Examples](https://github.com/pytorch/examples)
* [PyTorch Models](https://pytorch.org/hub/)
* [Intro to Deep Learning with PyTorch from Udacity](https://www.udacity.com/course/deep-learning-pytorch--ud188)
* [Intro to Machine Learning with PyTorch from Udacity](https://www.udacity.com/course/intro-to-machine-learning-nanodegree--nd229)
* [Deep Neural Networks with PyTorch from Coursera](https://www.coursera.org/learn/deep-neural-networks-with-pytorch)
* [PyTorch Twitter](https://twitter.com/PyTorch)
* [PyTorch Blog](https://pytorch.org/blog/)
* [PyTorch YouTube](https://www.youtube.com/channel/UCWXI5YeOsh03QvJ59PMaXFw)
## Communication
* Forums: Discuss implementations, research, etc. https://discuss.pytorch.org
* GitHub Issues: Bug reports, feature requests, install issues, RFCs, thoughts, etc. See the [issue lifecycle](CONTRIBUTING.md#issue-lifecycle) for how they are handled.
* Slack: The [PyTorch Slack](https://pytorch.slack.com/) hosts a primary audience of moderate to experienced PyTorch users and developers for general chat, online discussions, collaboration, etc.
* Facebook Page: Important announcements about PyTorch. https://www.facebook.com/pytorch
* For brand guidelines, please visit our website at [pytorch.org](https://pytorch.org/)
## Releases and Contributing
To report issues or contribute changes to PyTorch, see [CONTRIBUTING.md](CONTRIBUTING.md).
For information about PyTorch releases, see [RELEASE.md](RELEASE.md).
## The Team
PyTorch is a community-driven project with several skillful [engineers and researchers](https://docs.pytorch.org/docs/stable/community/persons_of_interest.html) contributing to it.
Note: This project is unrelated to [hughperkins/pytorch](https://github.com/hughperkins/pytorch) with the same name. Hugh is a valuable contributor to the Torch community and has helped with many things Torch and PyTorch.
## License
PyTorch has a BSD-style license, as found in the [LICENSE](LICENSE) file.