Duplicate cat if sandwiched between deq/quant (#19925)#19925
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19925
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 15 PendingAs of commit 7bd12f9 with merge base b63adec ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
Contributor
|
@DrJessop has exported this pull request. If you are a Meta employee, you can view the originating Diff in D107174424. |
This PR needs a
|
meta-codesync
Bot
changed the title
DrJessop
force-pushed
the
export-D107174424
branch
from
1f11fca to
c826de9
Compare
DrJessop
added a commit
to DrJessop/executorch
that referenced
this pull request
Jun 2, 2026
Summary:
Suppose we have (int_data -> dq, fp32_data) -> cat -> (slice, q)
This is a case with a multi-user cat. If we duplicated the cat, we would have
(int_data -> dq, fp32_data) -> (cat -> slice, cat -> q)
Then, in one of our chains, we could push the q above its cat. This allows for us to at least keep one chain using quantized math the whole way instead of forcing a break to fp32 in both chains.
Example {F1990607611}. Just suppose that we already hoisted the quant node to be right underneath the cat.
In a later pass, will hoist quant under single-user cat above the cat. This pass only duplicates the cat for each quant op user which has independent quant params.
Reviewed By: mcremon-meta
Differential Revision: D107174424
DrJessop
added a commit
to DrJessop/executorch
that referenced
this pull request
Jun 2, 2026
Summary:
Suppose we have (int_data -> dq, fp32_data) -> cat -> (slice, q)
This is a case with a multi-user cat. If we duplicated the cat, we would have
(int_data -> dq, fp32_data) -> (cat -> slice, cat -> q)
Then, in one of our chains, we could push the q above its cat. This allows for us to at least keep one chain using quantized math the whole way instead of forcing a break to fp32 in both chains.
Example {F1990607611}. Just suppose that we already hoisted the quant node to be right underneath the cat.
In a later pass, will hoist quant under single-user cat above the cat. This pass only duplicates the cat for each quant op user which has independent quant params.
Reviewed By: mcremon-meta
Differential Revision: D107174424
DrJessop
force-pushed
the
export-D107174424
branch
from
c826de9 to
56bfbdb
Compare
aliafzal
approved these changes
Jun 2, 2026
DrJessop
added a commit
to DrJessop/executorch
that referenced
this pull request
Jun 2, 2026
Summary:
Suppose we have (int_data -> dq, fp32_data) -> cat -> (slice, q)
This is a case with a multi-user cat. If we duplicated the cat, we would have
(int_data -> dq, fp32_data) -> (cat -> slice, cat -> q)
Then, in one of our chains, we could push the q above its cat. This allows for us to at least keep one chain using quantized math the whole way instead of forcing a break to fp32 in both chains.
Example {F1990607611}. Just suppose that we already hoisted the quant node to be right underneath the cat.
In a later pass, will hoist quant under single-user cat above the cat. This pass only duplicates the cat for each quant op user which has independent quant params.
Reviewed By: mcremon-meta
Differential Revision: D107174424
DrJessop
added a commit
to DrJessop/executorch
that referenced
this pull request
Jun 2, 2026
Summary:
Suppose we have (int_data -> dq, fp32_data) -> cat -> (slice, q)
This is a case with a multi-user cat. If we duplicated the cat, we would have
(int_data -> dq, fp32_data) -> (cat -> slice, cat -> q)
Then, in one of our chains, we could push the q above its cat. This allows for us to at least keep one chain using quantized math the whole way instead of forcing a break to fp32 in both chains.
Example {F1990607611}. Just suppose that we already hoisted the quant node to be right underneath the cat.
In a later pass, will hoist quant under single-user cat above the cat. This pass only duplicates the cat for each quant op user which has independent quant params.
Reviewed By: mcremon-meta
Differential Revision: D107174424
Summary:
Suppose we have (int_data -> dq, fp32_data) -> cat -> (slice, q)
This is a case with a multi-user cat. If we duplicated the cat, we would have
(int_data -> dq, fp32_data) -> (cat -> slice, cat -> q)
Then, in one of our chains, we could push the q above its cat. This allows for us to at least keep one chain using quantized math the whole way instead of forcing a break to fp32 in both chains.
Example {F1990607611}. Just suppose that we already hoisted the quant node to be right underneath the cat.
In a later pass, will hoist quant under single-user cat above the cat. This pass only duplicates the cat for each quant op user which has independent quant params.
Reviewed By: mcremon-meta
Differential Revision: D107174424
DrJessop
force-pushed
the
export-D107174424
branch
from
56bfbdb to
7bd12f9
Compare
Gasoonjia
pushed a commit
that referenced
this pull request
Jun 8, 2026
Differential Revision: D107174424 Pull Request resolved: #19925
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.
Summary:
Suppose we have (int_data -> dq, fp32_data) -> cat -> (slice, q)
This is a case with a multi-user cat. If we duplicated the cat, we would have
(int_data -> dq, fp32_data) -> (cat -> slice, cat -> q)
Then, in one of our chains, we could push the q above its cat. This allows for us to at least keep one chain using quantized math the whole way instead of forcing a break to fp32 in both chains.
Example {F1990607611}. Just suppose that we already hoisted the quant node to be right underneath the cat.
In a later pass, will hoist quant under single-user cat above the cat. This pass only duplicates the cat for each quant op user which has independent quant params.
Reviewed By: mcremon-meta
Differential Revision: D107174424