How to Install `torch` in Python

v2.11.0 Machine Learning & AI Python >=3.10 BSD-3-Clause

Tensors and Dynamic neural networks in Python with strong GPU acceleration

Install pip install torch

What is torch?

Tensors and Dynamic neural networks in Python with strong GPU acceleration

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.

PyPI page → Author: PyTorch Team <packages@pytorch.org>

Quick Start

Minimal example to get started with torch:

import torch

device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using: {device}")

x = torch.tensor([1.0, 2.0, 3.0]).to(device)
print(x.mean())   # tensor(2.)

Installation

pip (standard)

pip install torch

Virtual environment (recommended)

python -m venv venv
source venv/bin/activate   # Windows: venv\Scripts\activate
pip install torch

pip3

pip3 install torch

conda

conda install -c conda-forge torch

Poetry

poetry add torch

Dependencies

Installing torch will also install these packages:

filelock typing-extensions setuptools sympy networkx jinja2 fsspec cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx] cuda-bindings nvidia-cudnn-cu13 nvidia-cusparselt-cu13 nvidia-nccl-cu13

Verify the Installation

After installing, confirm the package is available:

python -c "import torch; print(torch.__version__)"

If this prints a version number, installation succeeded. If you see a ModuleNotFoundError, see the errors section below.

Installation Errors

Common errors when installing torch with pip.

ModuleNotFoundError: No module named 'torch'

Cause: The package is not installed in the current Python environment.

Fix: Run pip install torch. If using a virtual environment, ensure it is activated first.

ModuleNotFoundError: No module named 'torch' (installed but still failing)

Cause: pip installed the package into a different Python than the one running your script.

Fix: Use python -m pip install torch to install into the interpreter you are running.

ImportError: cannot import name 'X' from 'torch'

Cause: The function or class does not exist in the installed version.

Fix: Check the version with pip show torch and upgrade with pip install --upgrade torch.

pip: command not found

Cause: pip is not in PATH or Python was not added to PATH during installation.

Fix: Try python -m pip install torch. On macOS/Linux try pip3.

PermissionError: [Errno 13] Permission denied

Cause: No write access to the system Python package directory.

Fix: Use a virtual environment, or add --user: pip install --user torch

SSL: CERTIFICATE_VERIFY_FAILED

Cause: pip cannot verify PyPI's SSL certificate — common behind corporate proxies.

Fix: Try: pip install --trusted-host pypi.org --trusted-host files.pythonhosted.org torch

RuntimeError: CUDA out of memory

Cause: Model or batch size exceeds GPU memory.

Fix: Reduce batch size, call torch.cuda.empty_cache(), or run on CPU.

Recent Releases

Version	Released
`2.11.0` latest	2026-03-23
`2.10.0`	2026-01-21
`2.9.1`	2025-11-12
`2.9.0`	2025-10-15
`2.8.0`	2025-08-06

Full release history on PyPI →

Manage torch

Upgrade to latest version

pip install --upgrade torch

Install a specific version

pip install torch==2.11.0

Uninstall

pip uninstall torch

Check what is installed

pip show torch

Last updated: 2026-04-11 • Data from PyPI

How to Install torch in Python