Python teams struggle to install CPU-only PyTorch cleanly
Python developers who need PyTorch without GPU support still face large downloads and a separate CPU-wheel installation path. That path is awkward to use with tools such as Poetry and internal package mirrors, where version-string handling can cause additional problems. Some developers say the wheel experience has put them off using PyTorch.
For python developers and application maintainers deploying CPU-only PyTorch. Mentioned from Sep 2019 to Aug 2025 on GitHub and Hacker News.
4 different people described this problem in 2 separate discussions.
- Indie fit
- 3.0/10
- Pain
- 5.5/10
- Frequency
- 5.8/10
- Willingness to pay
- 0.0/10
- Momentum
- 5.0/10
- Who pays
- Professionals
- Competition
- Medium
- Build difficulty
- Medium
What people said
Quoted word for word. Follow a link to read the whole discussion.
We're using torch in https://github.com/neuropoly/spinalcordtoolbox/ but we can't ask our users to download almost a gigabyte of software. Not everyone is running in a big high performance data centre
Using PyTorch in production does not require necessarily the (~700 MB big) GPU version, but installing the much smaller CPU version as suggested on the website: makes it hard to use tools like Poetry, which do not work with pip itself and therefore do not support an argument like -f https://download.pytorch.org/whl/tor…
Build brief
See what to build and who will buy it
- 2 product ideas with the smallest useful version and pricing
- 4 places to find your first customers
- 2 more quotes from people who have this problem
- Current workarounds, existing solutions and risks