pytorch-wheel-installer 0.1.1.post0

Install PyTorch distributions from the latest wheels.

Disclaimer

Neither this project (pytorch_wheel_installer) nor its author (Philip Meier) are affiliated with PyTorch in any way. PyTorch and any related marks are trademarks of Facebook, Inc.

pytorch_wheel_installer

Commandline utility and tox -plugin to install PyTorch distributions from the latest wheels. The computation backend (CPU, CUDA), the language version, and the platform are detected automatically but can be overwritten manually.

package
code
tests

Installation

The latest stable version can be installed with

pip install pytorch_wheel_installer

The latest potentially unstable version can be installed with

pip install git+https://github.com/pmeier/pytorch_wheel_installer

Usage

CLI

The CLI can be invoked by pytorch_wheel_installer or its shorthand pwi.

$ pwi --help
usage: pwi [-h] [--version] [--distribution DISTRIBUTION] [--backend BACKEND]
           [--language LANGUAGE] [--platform PLATFORM] [--no-install]
           [--pip-cmd PIP_CMD]

Install PyTorch from the latest wheels.

optional arguments:
  -h, --help            show this help message and exit
  --version, -v         Show version and exit.
  --distribution DISTRIBUTION, -d DISTRIBUTION
                        PyTorch distribution e.g. 'torch', 'torchvision'.
                        Multiple distributions can be given as a comma-
                        separated list. Defaults to 'torch,torchvision'.
  --backend BACKEND, -b BACKEND
                        Computation backend e.g. 'cpu' or 'cu102'. If not
                        given the backend is automatically detected from the
                        available hardware preferring CUDA over CPU.
  --language LANGUAGE, -l LANGUAGE
                        Language implementation and version tag e.g. 'py3',
                        'cp36'. Defaults to the language version used to run
                        this.
  --platform PLATFORM, -p PLATFORM
                        Platform e.g. 'linux', 'windows', 'macos', or 'any'.
                        Defaults to the platform that is used to run this.
  --no-install, -ni     If given, the selected wheels are written to STDOUT
                        instead of installed.
  --pip-cmd PIP_CMD, -pc PIP_CMD
                        pip command that is used to install the wheels.
                        Defaults to 'pip install'

The --no-install option can be used to pipe or redirect the wheel links such as generating a requirements.txt file:

$ pwi --no-install > requirements.txt
$ cat requirements.txt
https://download.pytorch.org/whl/cu102/torch-1.5.1-cp36-cp36m-linux_x86_64.whl
https://download.pytorch.org/whl/cu102/torchvision-0.6.1-cp36-cp36m-linux_x86_64.whl

tox

$ tox --help
...
optional arguments:
...
--pytorch-install                Install PyTorch distributions from the latest
                                 wheels before the other dependencies. (default:
                                 False)
--pytorch-distribution DISTRIBUTION
                                 PyTorch distribution e.g. 'torch', 'torchvision'.
                                 Multiple distributions can be given as a
                                 comma-separated list. Defaults to
                                 'torch,torchvision'. (default:torch,torchvision)
--pytorch-backend BACKEND        Computation backend e.g. 'cpu' or 'cu102'. If not
                                 given the backend is automatically detected from the
                                 available hardware preferring CUDA over CPU.
                                 (default: None)
--pytorch-language LANGUAGE      Language implementation and version tag e.g. 'py3',
                                 'cp36'. Defaults to the language version used to run
                                 this. (default: None)
--pytorch-platform PLATFORM      Platform e.g. 'linux', 'windows', 'macos', or 'any'.
                                 Defaults to the platform that is used to run this.
                                 (default: None)
...

If --pytorch-install is not given, nothing is installed.