Closed jiaxiyan closed 3 months ago
This reverts commit 6aa6708f99234ddef233182d656ec25eb1c5159b because NCCL alltoall test fails to register memory with dmabuf on 16 nodes.
Please include the reason of the revert in the commit message
Is the full revert necessary? Or is this just a temporary mitigation until we know more?
This reverts commit 6aa6708f99234ddef233182d656ec25eb1c5159b because NCCL alltoall test fails to register memory with dmabuf on 16 nodes.