Count a tensor as source of factor axes candidate if the factor sharding of the tensor is a strict prefix of the candidate axes because dynamic-slice is free. #888
+131
−136
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Count a tensor as source of factor axes candidate if the factor sharding of the tensor is a strict prefix of the candidate axes because dynamic-slice is free.
In addition to that a tensor is already counted as source of factor axes candidate if the factor sharding of the tensor is a (not necessarily strict) prefix of the candidate axes.
For example; Given [i//{x}] + [i//{x, y}] = [i//{x, y, z}], FactorAxesCandidate (i// {x,y}) has the following sources:
BEFORE: LHS, RHS
AFTER: LHS, RHS, RESULT
Given [i][k//{x}] @ [k][j//{x}] = [i//{x}][j],
BEFORE:
AFTER:
It implies that the largest factor, in this particular case, wins {x}.