[IMPORTANT] End-of-Life Notification for oneCCL Bindings for PyTorch* (torch-ccl)

[tye1]

Dear Community,

We launched oneCCL Bindings for PyTorch* (torch-ccl) in 2020 to enable distributed communication support together with Intel® Extension for PyTorch* for Intel® CPU and GPU platforms. Over the years, distributed support for Intel® platforms has now been successfully upstreamed into PyTorch* itself.

As a result, we have discontinued active development of torch-ccl following the 2.8 release. We will continue to provide critical bug fixes and security patches throughout 2025 to ensure a smooth transition for our partners and the broader community. We strongly recommend using PyTorch* directly for distributed scenarios going forward, as we remain committed to delivering robust support and performance with PyTorch* for Intel® CPU and GPU platforms.

If you have any questions or concerns, please feel free to comment here.
Thank you all for your continued support!

[tye1]

Since we have not received any concerns from our partners or the community so far, we plan to archive this repository by the end of 2025. Please feel free to comment if you have any objections.