Skip to main content

This is a meta package to install PyTorch for different accelerator versions

Project description

What is this?

This repository is a proof of concept on how we can change PyTorch's package installation for the better. It should enable a path to where we can have users specify which specific accelerator they would like to use by using the following installation command(s):

[!NOTE] This doesn't work as of today this is just showcasing what this could look like if we rolled this as the default torch package.

# Based on accelerator type
pip install torch[cpu]
pip install torch[cuda]
pip install torch[rocm]
pip install torch[xpu]

# Based on accelerator version
pip install torch[cpu-cxx11-abi]
pip install torch[cuda12-4]
pip install torch[cuda12-1]
pip install torch[rocm6-2]
pip install torch[rocm6-1]

How to build

Pre-requisites:

  • uv
  • python3.12 (with build module installed)

Commands to run

# Build the wheel and rename to platform specific
./generate.py

Work Todo

  • Create automation to build on Linux, macOS, & Windows through Github Actions
  • Automate this to automatically reflect PyTorch's binary build matrix

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

torchpick-0.0.1-py3-none-any.whl (1.7 kB view details)

Uploaded Python 3

File details

Details for the file torchpick-0.0.1-py3-none-any.whl.

File metadata

  • Download URL: torchpick-0.0.1-py3-none-any.whl
  • Upload date:
  • Size: 1.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.6

File hashes

Hashes for torchpick-0.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 bb4ce80ea9016c65235f53fb1c20bb1508068f603a0115596e2f28204e1917be
MD5 eb1b170177e77a1e980f1229acbac331
BLAKE2b-256 e62d1325a706208f306251dff27302569e4ff603c8cabe93ae3f270f3aa946bc

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page