[QNN EP] Improve QDQ model accuracy tests #16916
adrianlizarraga
merged
35 commits into
main
from
adrianl/qnn-add-unit-tests
Add tests
caed9a1d
Fix merge conflicts from main
4f22f43a
Rework new Asin, Sin, and Sign tests to use float inputs/outputs
afbe4677
Add 3-way comparision between Cpu(f32), Cpu(QDQ), QNN(QDQ)
59948ed2
Clean up
35b91ee6
Switch argmin/argmax tests to new QDQ accuracy testing
eee35dbe
Convert averagepool tests (need to disable inaccurate tests)
fa5dcb43
Add explicit inputs to AvergePool tests
aeba185f
Use new QDQ acc testing for BatchNorm. Found issues.
9fcba0b4
Update conv tests to new accuracy testing
b6123fde
Update comments for QDQ accuracy testing func
ec41d97f
Fix linter errors. Add include for math
d89e0b36
More linter fixes. Use difference cmath funcs
40ac48a4
Update shape for And/Or tests
a610a6ee
More lint error
1eed952a
Update Gather op tests. Need to support testing scalar indices
9c6149f9
Simplify accuracy computation
46f47933
Convert InstanceNorm to new testing approach
7de0d078
Update LeakyRelu tests to new accuracy testing approach
974bc774
Update LRN tests
f91aea14
Update MatMul tests
d48980fc
Fix merge conflicts with main
2a055548
Reuse GetDataQuantParams function
f82fb22e
Update MaxPool tests
e513d631
Update ReduceOp cpu tests
fd01af35
Update QDQ Reduce op tests
56fa59be
Explicitly handle quant params when data only has a single 0
7af94106
Update Resize tests to use new accuracy measuring approach
962bca6b
HectorSVC
commented
on 2023-07-31
Add disabled inaccurate tests
7917a070
Add a way to override a test input's range
f2216ee3
Move all reduce op tests into 1 file and rename it
32ea966b
Merge latest commits from main
0742fd3c
HectorSVC
commented
on 2023-08-03
HectorSVC
commented
on 2023-08-03
HectorSVC
dismissed these changes
on 2023-08-03
Merge latest commits from main
20edf005
Revert change in input shape
3878e79c
Merge latest commits from main
a21251eb
HectorSVC
approved these changes
on 2023-08-03
Login to write a write a comment.
Login via GitHub