Lunderberg commented on code in PR #16098:
URL: https://github.com/apache/tvm/pull/16098#discussion_r1391647705
##########
python/tvm/relax/op/distributed/distributed.py:
##########
@@ -59,3 +59,28 @@ def redistribute(input: Expr, device_mesh: DeviceMesh,
placement: Placement) ->
The tensor after redistribution.
"""
return _ffi_api.redistribute(input, device_mesh, placement) # type: ignore
+
+
+def redistribute_replica_to_shard(input: Expr, num_workers: int, axis: int) ->
Expr:
+ """Slice tensor into several parts along one axis,
+ and each worker takes one part.
+ Assumes input is already broadcasted.
+ This is a specialized version of redistribute op.
+
+ Parameters
+ ----------
+ input : relax.Expr
+ The buffer to be sliced into equal parts.
+
+ num_worker : int
Review Comment:
Thank you, and the docstring update helps make the requirement clearer.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
