PR #5524 GatherGrad optimization

GatherGrad optimization #5524

edgchen1 merged 20 commits into master from edgchen1/gather_grad_optimization

GatherGrad alternate implementation.

7a07178d

Merge remote-tracking branch 'origin/master' into edgchen1/gather_gra…

48453d57

Add PartialSums implementation.

19dab862

Merge remote-tracking branch 'origin/master' into edgchen1/gather_gra…

7f35ec65

Fix build errors.

7474ddb0

Make test handle previously set environment variables.

90863a4f

Update GatherGrad impl.

7631a91c

Merge remote-tracking branch 'origin/master' into edgchen1/gather_gra…

3f388241

Renaming.

bf748454

Update threshold, cleanup.

e076c3f3

Use specific cub headers.

62fe5d61

Update GatherGrad tests.

b7950a8d

Remove --expt-extended-lambda nvcc option.

e80c97b5

Revert change to debug node IO test.

e042dfc4

edgchen1 added type:performance

edgchen1 added training

edgchen1 added ep:CUDA

edgchen1 requested a review from

suffiank 5 years ago

edgchen1 requested a review from

ytaous 5 years ago

edgchen1 requested a review from

SherlockNoMad 5 years ago

edgchen1 requested a review 5 years ago

Fix build issues.

58f5c797

Update tests.

ba9dedfe

Update GatherGrad tests.

83003039

Fix build.

bae564b3

Fix build error.

af874ec4

SherlockNoMad dismissed these changes on 2020-10-22

ytaous dismissed these changes on 2020-10-22

weixingzhang commented on 2020-10-22

Reuse AccumulationType in other places.

3b305818

edgchen1 dismissed their stale review via 3b305818 5 years ago

weixingzhang approved these changes on 2020-10-22

edgchen1 closed this 5 years ago

edgchen1 reopened this 5 years ago

edgchen1 merged 68fe7226 into master 5 years ago

edgchen1 deleted the edgchen1/gather_grad_optimization branch 5 years ago

Reviewers

weixingzhang

ytaous

SherlockNoMad

suffiank

Assignees

No one assigned

Labels

training ep:CUDA

Milestone

No milestone

onnxruntime GatherGrad optimization #5524 Merged

GatherGrad optimization #5524

onnxruntime
GatherGrad optimization
#5524

Merged