Deprecate tag qdq pass in xnnbackend #12170
Merged
Conversation
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/12170
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 19 PendingAs of commit c46b635 with merge base a909b83 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D77055623 |
mcr229
approved these changes
Jul 2, 2025
abhinaykukkadapu
force-pushed
the
export-D77055623
branch
from
b9d8cca to
85e7b8c
Compare
|
@abhinaykukkadapu has imported this pull request. If you are a Meta employee, you can view this in D77055623. |
abhinaykukkadapu
added a commit
to abhinaykukkadapu/executorch
that referenced
this pull request
Jul 2, 2025
Summary:
This diff decentralizes the q/dq implicit node tagging to individual partition configs instead of tagging it as part of backend pass.
Changes in this diff:
1. Deprecate tag q dq pass
2. Remove all the places where this pass is used in the backend preprocess phase.
3. Decentralize the tagging to individual configs
a. `generic_node_configs` will handle most of the non gemm nodes
b. `gemm_configs` will handle gemm nodes
c. channels last pass will add (copy q dq) or (dq copy q), tag the relevant nodes.
d. tag q dq in conv1d unsqueeze pass.
e. Tag q dq in compose cat
4. Deprecate configs.py where all the collection of nodes is maintained
Fixes: pytorch#11588
Test Plan:
Imported from GitHub, without a `Test Plan:` line.
Rollback Plan:
Differential Revision: D77055623
Pulled By: abhinaykukkadapu
abhinaykukkadapu
force-pushed
the
export-D77055623
branch
from
85e7b8c to
76418fd
Compare
|
This pull request was exported from Phabricator. Differential Revision: D77055623 |
Summary:
This diff decentralizes the q/dq implicit node tagging to individual partition configs instead of tagging it as part of backend pass.
Changes in this diff:
1. Deprecate tag q dq pass
2. Remove all the places where this pass is used in the backend preprocess phase.
3. Decentralize the tagging to individual configs
a. `generic_node_configs` will handle most of the non gemm nodes
b. `gemm_configs` will handle gemm nodes
c. channels last pass will add (copy q dq) or (dq copy q), tag the relevant nodes.
d. tag q dq in conv1d unsqueeze pass.
e. Tag q dq in compose cat
4. Deprecate configs.py where all the collection of nodes is maintained
Fixes: pytorch#11588
Test Plan:
Imported from GitHub, without a `Test Plan:` line.
Rollback Plan:
Differential Revision: D77055623
Pulled By: abhinaykukkadapu
abhinaykukkadapu
force-pushed
the
export-D77055623
branch
from
76418fd to
c46b635
Compare
|
@abhinaykukkadapu has imported this pull request. If you are a Meta employee, you can view this in D77055623. |
|
@pytorchbot label "release notes: none" |
pytorch-bot
bot
added
the
release notes: none
Tanish2101
pushed a commit
to Tanish2101/executorch
that referenced
this pull request
Jul 9, 2025
Summary:
This diff decentralizes the q/dq implicit node tagging to individual
partition configs instead of tagging it as part of backend pass.
Changes in this diff:
1. Deprecate tag q dq pass
2. Remove all the places where this pass is used in the backend
preprocess phase.
3. Decentralize the tagging to individual configs
a. `generic_node_configs` will handle most of the non gemm nodes
b. `gemm_configs` will handle gemm nodes
c. channels last pass will add (copy q dq) or (dq copy q), tag the
relevant nodes.
d. tag q dq in conv1d unsqueeze pass.
e. Tag q dq in compose cat
4. Deprecate configs.py where all the collection of nodes is maintained
Fixes: pytorch#11588
Reviewed By: mcr229
Differential Revision: D77055623
digantdesai
reviewed
Jul 11, 2025
| q.meta = copy.meta.copy() | ||
|
|
||
| # Always tag q/dq as implicit | ||
| tag_as_implicit_q_dq(dq) |
There was a problem hiding this comment.
I was talking with @mcr229 about this, this can be brittle i.e. implicit tagging for q/dq in >1 places. Do we want to add a pass in the future to verify if the tagging is correct or not. Just for validation? Feel free to create an issue, don't have to add it right away.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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:
This diff decentralizes the q/dq implicit node tagging to individual partition configs instead of tagging it as part of backend pass.
Changes in this diff:
a.
generic_node_configswill handle most of the non gemm nodesb.
gemm_configswill handle gemm nodesc. channels last pass will add (copy q dq) or (dq copy q), tag the relevant nodes.
d. tag q dq in conv1d unsqueeze pass.
e. Tag q dq in compose cat
Fixes: #11588
Reviewed By: mcr229
Differential Revision: D77055623