supporting dynamo compilation of end-to-end send/recv in distributed #153642
Labels
module: dynamo
oncall: distributed
Add this issue/PR to distributed oncall triage queue
oncall: pt2
triaged
This issue has been looked at a team member, and triaged and prioritized into an appropriate module
🚀 The feature, motivation and pitch
Currently the send and receive operations in torch distributed are not supported by the dynamo compiler, in that the compilation would fail for user-defined functions that attempt to perform these end-to-end communications. However, these operations are important to attain high performance for some distributed hardware architectures/platforms.
We are working on extending the dynamo compilation to support this new feature. We would like to upstream the extension to the main branch of the repository.
Alternatives
No response
Additional context
No response
cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @amjames
The text was updated successfully, but these errors were encountered: