pytorch

7a15e823 - Fix tensor registration to work with coalescing collectives. (#99763)

Fix tensor registration to work with coalescing collectives. (#99763) We do it by making it possible to register multiple tensors for the same worker and coordinate waiting/cleanup among them. This ensures waiting on any number the output tensors will result in a single stream sync. This simplifies codegen by inductor. Pull Request resolved: https://github.com/pytorch/pytorch/pull/99763 Approved by: https://github.com/wanchaol