Use upstream DTensor convert strategy#495
Closed
aditvenk wants to merge 1 commit into
Closed
Conversation
aditvenk
force-pushed
the
aditvenk/stack/7
branch
from
9cb11ee to
68fe2d8
Compare
aditvenk
force-pushed
the
aditvenk/stack/6
branch
from
39543fd to
2b69c38
Compare
This was referenced Jun 15, 2026
meta-cla
Bot
added
the
CLA Signed
This was referenced Jun 15, 2026
fmassa
reviewed
Jun 15, 2026
fmassa
left a comment
fmassa
left a comment
Contributor
There was a problem hiding this comment.
Can you share pointers to where the convert_element_type rule is defined in PyTorch? I couldn't find it
PyTorch now provides a single-dim strategy for prims.convert_element_type.default, so AutoParallel only needs to keep the local propagation rule for its custom dtype_cast op. Authored with Claude. stack-info: PR: #495, branch: aditvenk/stack/7
aditvenk
force-pushed
the
aditvenk/stack/7
branch
from
68fe2d8 to
f363b32
Compare
Contributor
Author
|
closing in favor of #505 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Stacked PRs:
Use upstream DTensor convert strategy
PyTorch now provides a single-dim strategy for prims.convert_element_type.default, so AutoParallel only needs to keep the local propagation rule for its custom dtype_cast op.
Authored with Claude.