-
Notifications
You must be signed in to change notification settings - Fork 4.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[p2p]add object list send/recv #1024
[p2p]add object list send/recv #1024
Conversation
This reverts commit df7e650.
colossalai/communication/p2p.py
Outdated
for tensor_send in object_send_next: | ||
send_next_op = dist.P2POp(dist.isend, tensor_send, next_rank) | ||
ops.append(send_next_op) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Torch has this API torch.distributed.broadcast_object_list
, not sure how the performance is like compared to a list of P2P Op.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The API broadcast_object_list sends the data to all ranks in group. I think this pr is something like send/recv_object_list.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
One way to do this is to create a process group for each pair of adjacent stages, and broadcast from one rank to the other rank in this process group.
No description provided.