Hi Everyone, we from PyTorch Release Engineering team are actively working on PyTorch 2.0 release (see the milestone) and wish to call out the support change concerning CUDA and Python. For Release 2.0 we target to support following CUDA and Python configurations:
- CUDA 11.7 as stable version and CUDA 11.8 as experimental version of CUDA. Dropping support for CUDA 11.6
- Python >=3.8, <=3.11 . Dropping Support for Python 3.7
If you still are using/depending on CUDA 11.6 or Python 3.7 builds we strongly recommend moving to at least CUDA 11.7 and Python 3.8, as it would be the minimum version required for PyTorch 2.0.
Please refer to Release Compatibility Matrix here:
https://github.com/pytorch/pytorch/blob/master/RELEASE.md#release-compatibility-matrix