Simple utility to install pytorch, pytorch-geometric and pytorch-lightning. Detects CUDA version automatically.
Project description
PyTorch Install Helper
Super simple helper to install pytorch stuff without having to check cuda versions and go to websites for the installer URLs. Installs based on requested or detected CUDA version, doesn't check python versions.
Only Linux and macOS Supported
Detects what cuda version is available and runs the pip command to install latest pytorch and compatible cuda version
Installation
pip install torchinstaller
Usage
$ torchinstall -h
usage: torchinstall [-h] [--pytorch [PYTORCH]] [--pyg] [--pyg-lib-source] [--cuda {cu102,cu111,cu113,cu116,cu117,cu118,rocm4.0.1,rocm4.1,rocm4.2,rocm4.5.2,rocm5.1.1,rocm5.2,rocm5.4.2}] [--lightning] [--use {pip,conda,mamba,poetry}] [-install]
options:
-h, --help show this help message and exit
--pytorch [PYTORCH], -pt [PYTORCH]
Flag to install pytorch, can optionally specify a desired version. Must be full semantic version, e.g. 1.13.1, not 1.13, or `latest`
--pyg, -pyg Flag to install pytorch-geometric
--pyg-lib-source, -pyg-src
Flag to install PyG from source. i.e. PyG doesn't support wheels for M1/M2 macs. They recommend installing from source
--cuda {cu102,cu111,cu113,cu116,cu117,cu118,rocm4.0.1,rocm4.1,rocm4.2,rocm4.5.2,rocm5.1.1,rocm5.2,rocm5.4.2}, -c {cu102,cu111,cu113,cu116,cu117,cu118,rocm4.0.1,rocm4.1,rocm4.2,rocm4.5.2,rocm5.1.1,rocm5.2,rocm5.4.2}
Manually specify platform version (cuda or rocm) instead ofauto-detect (useful for cluster installations).
--lightning, -l Flag to install pytorch-lightning
--use {pip,conda,mamba,poetry}, -u {pip,conda,mamba,poetry}
set command to install with.
-install, -i Run installation (default is to dry run commands)
Note: pytorch-geometric can be problematic to install. Installing from source has been added to facilitate installation, but referring to their documentation may be necessary to address errors if they occur.
Note: poetry support is a work in-progress and is unstable
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for torchinstaller-0.5.6-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | b431055959db5ed43682d69a9ec0e78d209bfa9356bf91ea2b3f1ea782cafba6 |
|
MD5 | a809d633f6a478a14faa1bbf3626bbbc |
|
BLAKE2b-256 | d700315d9b4eed2cb18a0782771038cee67122f73acd8b79f871a99ec4416114 |