onnxruntime
Graph transformer to ensure unique DQ nodes for QDQ node units
#15145
Merged

Graph transformer to ensure unique DQ nodes for QDQ node units #15145

edgchen1
edgchen1 initial pass at EnsureUniqueDQForNodeUnit
ce3f80a2
edgchen1 refine implementation
c8f71467
edgchen1 minor update
33c7ae18
edgchen1 save work
c90d50b3
edgchen1 continue working on impl and tests
859d201e
edgchen1 refine tests, update graph to account for manually set output that is…
a90a4a25
edgchen1 Merge remote-tracking branch 'origin/main' into edgchen1/dq_duplicati…
12d3d955
edgchen1 fix tests and tidy up
bde8b5a6
edgchen1 update class documentation comment
972958f0
edgchen1 edgchen1 requested a review from skottmckay skottmckay 3 years ago
edgchen1 edgchen1 requested a review from wejoncy wejoncy 3 years ago
edgchen1 edgchen1 requested a review from yufenglee yufenglee 3 years ago
edgchen1 edgchen1 requested a review from YUNQIUGUO YUNQIUGUO 3 years ago
edgchen1
edgchen1 commented on 2023-03-21
edgchen1 address warning, add include
43b5381c
edgchen1 fix ORT_RETURN_IF_NOT
e95fe0d9
edgchen1 reorder optimizers so transpose optimizer is last
e18f483e
edgchen1
edgchen1 commented on 2023-03-22
edgchen1
edgchen1 fix ORT_RETURN_IF_NOT
86308d92
wejoncy
wejoncy commented on 2023-03-22
HectorSVC
edgchen1 make constexpr
fe61a3b5
edgchen1 Add test that uses model.
3621db22
edgchen1
skottmckay
skottmckay commented on 2023-03-23
skottmckay
HectorSVC
edgchen1
edgchen1 Set inlined vector size to fix build on Windows.
d48ed2f1
edgchen1 Move DQ validity check to NodeUnit construction.
c824b3e9
edgchen1 Refine comments and variable names.
c3aa98a9
edgchen1 Remove python DQ duplication implementation.
52542736
edgchen1 make EnsureUniqueDQForNodeUnit transformer required, update tests
5f956c31
edgchen1
edgchen1 commented on 2023-03-24
edgchen1
edgchen1 commented on 2023-03-24
edgchen1
edgchen1 commented on 2023-03-24
edgchen1
edgchen1 commented on 2023-03-24
edgchen1
edgchen1 commented on 2023-03-24
edgchen1 Merge remote-tracking branch 'origin/main' into edgchen1/dq_duplicati…
30d69f02
edgchen1 Address comments, fix test.
d22743c7
edgchen1 fix no contrib ops build
4e23f5a1
edgchen1
edgchen1 commented on 2023-03-27
edgchen1 refine
0ae9bc1e
edgchen1 add more space to comments
98ca2385
skottmckay
skottmckay commented on 2023-03-28
edgchen1 rename variable for clarity
0f67f0e6
edgchen1 Merge remote-tracking branch 'origin/main' into edgchen1/dq_duplicati…
337e4ba0
yuslepukhin
yuslepukhin commented on 2023-03-28
edgchen1
edgchen1 commented on 2023-03-28
edgchen1 use constant for inlined vector default size element size cutoff
4299226e
edgchen1 use inference session to test
da762410
edgchen1
edgchen1 commented on 2023-03-28
edgchen1 put valid DQ check in helper function, also run it when forming QDQ n…
7e110fb5
edgchen1 add DQ duplication transformer to level1 optimizer group
4c50e1d2
edgchen1 remove CheckDQNodes declaration
eeec0a81
edgchen1 Merge remote-tracking branch 'origin/main' into edgchen1/dq_duplicati…
c1264bb3
edgchen1 update comment in graph.cc
3f18c3b2
edgchen1 Add unit test from https://github.com/microsoft/onnxruntime/pull/15249.
b5b7f3ea
edgchen1 add note to onnxruntime\test\testdata\transform\qdq_with_multi_consum…
9421beb0
edgchen1 update graph.cc comment
df2b30eb
skottmckay
skottmckay dismissed these changes on 2023-03-30
edgchen1 add info to comment
421534ac
edgchen1 fix comparison warning
526373e1
edgchen1 update and re-enable transpose optimizer test
ed18eb48
edgchen1 edgchen1 dismissed their stale review via ed18eb48 3 years ago
skottmckay
skottmckay approved these changes on 2023-03-30
skottmckay skottmckay merged 9f942e1a into main 3 years ago
skottmckay skottmckay deleted the edgchen1/dq_duplication_transformer branch 3 years ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone