Skip to content

Added host_conv_wrw for verification#15

Merged
asroy merged 5 commits into
developfrom
add_host_conv_bwd_wrw
Aug 19, 2021
Merged

Added host_conv_wrw for verification#15
asroy merged 5 commits into
developfrom
add_host_conv_bwd_wrw

Conversation

@zjing14
Copy link
Copy Markdown
Contributor

@zjing14 zjing14 commented Aug 17, 2021 •
edited
Loading

  • Added host_direct_convolution_backward_weights function

@zjing14 zjing14 requested review from asleepzzz, asroy and ltqin August 17, 2021 14:35
ltqin
ltqin reviewed Aug 18, 2021
View reviewed changes
Comment thread host/host_tensor/include/host_conv_wrw.hpp
void host_direct_convolution_backward_weights(
const Tensor<TIn>& in,
const Tensor<TOut>& out,
Tensor<TWei>& wei,
Copy link
Copy Markdown
Collaborator

@ltqin ltqin Aug 18, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

in wei out or out in wei?

Copy link
Copy Markdown
Contributor Author

@zjing14 zjing14 Aug 18, 2021 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(in, out, wei). I should change the order to out, in, wei.

@ltqin
Copy link
Copy Markdown
Collaborator

ltqin commented Aug 18, 2021 •
edited
Loading

@zjing14 I didn't find out issue on code, but the result is not right.
branch: backward_weight_v4r4r2_xdlops
command: host/driver_offline/conv_wrw_driver_offline 0 0 1 1 0 1 128 256 128 3 3 14 14 1 1 1 1 0 0 0 0
in: dim 4, lengths {128, 128, 14, 14}, strides {25088, 196, 14, 1}
wei: dim 4, lengths {256, 128, 3, 3}, strides {1152, 9, 3, 1}
out: dim 4, lengths {128, 256, 12, 12}, strides {36864, 144, 12, 1}

device result: 18432
host result: 4608

@zjing14 zjing14 changed the title Add bwd and wrw into host conv for verification Added host_conv_wrw for verification Aug 18, 2021
@zjing14
Copy link
Copy Markdown
Contributor Author

zjing14 commented Aug 18, 2021

@zjing14 I didn't find out issue on code, but the result is not right.
branch: backward_weight_v4r4r2_xdlops
command: host/driver_offline/conv_wrw_driver_offline 0 0 1 1 0 1 128 256 128 3 3 14 14 1 1 1 1 0 0 0 0
in: dim 4, lengths {128, 128, 14, 14}, strides {25088, 196, 14, 1}
wei: dim 4, lengths {256, 128, 3, 3}, strides {1152, 9, 3, 1}
out: dim 4, lengths {128, 256, 12, 12}, strides {36864, 144, 12, 1}

device result: 18432
host result: 4608

There is a bug in the function. FIxed. Please pull and check.

@asroy
Copy link
Copy Markdown
Contributor

asroy commented Aug 18, 2021

@ltqin Please check if GPU and CPU result match

@ltqin ltqin self-requested a review August 19, 2021 01:10
@asroy asroy merged commit ba6f79a into develop Aug 19, 2021
@zjing14 zjing14 deleted the add_host_conv_bwd_wrw branch September 26, 2021 14:30
asroy pushed a commit that referenced this pull request Dec 1, 2023
@carlushuang
slice kv, and use 3d padding LDS layout (#15)
7b1a0b7 
* slice kv, and use 3d padding LDS layout

* add missing sync

* put sync to another poace

* move sync place

* revert to normal
asroy added a commit that referenced this pull request Dec 1, 2023
@asroy
Revert "slice kv, and use 3d padding LDS layout (#15)" (#18)
2dfbfbb 
This reverts commit 7b1a0b7.
@illsilin illsilin mentioned this pull request Oct 15, 2025
Merged
7 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@ltqin ltqin ltqin approved these changes

@asleepzzz asleepzzz Awaiting requested review from asleepzzz asleepzzz is a code owner

+1 more reviewer

@asroy asroy asroy approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@zjing14 @ltqin @asroy