Skip to content

Add aten mkldnn backward ops: max_pool2d, avg_pool2d and adaptive_av… #20571

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
XiaobingSuper wants to merge 2 commits into from
Closed

Add aten mkldnn backward ops: max_pool2d, avg_pool2d and adaptive_av… #20571

XiaobingSuper wants to merge 2 commits into from

Conversation

@XiaobingSuper
Copy link
Collaborator

@XiaobingSuper XiaobingSuper commented May 16, 2019 •
edited
Loading

mkldnn backward ops list:

@pytorchbot pytorchbot added module: autograd Related to torch.autograd, and the autograd engine in general module: internals Related to internal abstractions in c10 and ATen module: mkldnn Related to Intel IDEEP or oneDNN (a.k.a. mkldnn) integration module: operators labels May 16, 2019
This was referenced May 16, 2019
Closed
Closed
Closed
Closed
Closed
@li-roy li-roy added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label May 16, 2019
facebook-github-bot pushed a commit that referenced this pull request Jun 13, 2019
@XiaobingSuper @facebook-github-bot
Add mkldnn mul operator (#20575)
b599bb3 
Summary:
### mkldnn backward ops list:
 - [ ] \(#20567) Add aten mkldnn conv2d backward operator 💛
 - [ ] \(#20570) Add aten mkldnn backward ops: relu, linear and reshape 💛
 - [ ] \(#20571) Add aten mkldnn backward ops: max_pool2d, avg_pool2d and adaptive_avg_poo2d 💛
 - [ ] \(#20572) Add aten mkldnn batchnorm backward operator 💛
 - [ ] \(#20573) Add aten mkldnn zero_ operator:yellow_heart:
 - [ ] \(#20575) Add mkldnn mul operator 💛
Pull Request resolved: #20575

Differential Revision: D15799529

Pulled By: bddppq

fbshipit-source-id: 4887d8ef1a0e316ad9db199b657d9481fc13e486
zdevito pushed a commit to zdevito/ATen that referenced this pull request Jun 13, 2019
@XiaobingSuper @facebook-github-bot
Add mkldnn mul operator (#20575)
746db01 
Summary:
### mkldnn backward ops list:
 - [ ] \(pytorch/pytorch#20567) Add aten mkldnn conv2d backward operator 💛
 - [ ] \(pytorch/pytorch#20570) Add aten mkldnn backward ops: relu, linear and reshape 💛
 - [ ] \(pytorch/pytorch#20571) Add aten mkldnn backward ops: max_pool2d, avg_pool2d and adaptive_avg_poo2d 💛
 - [ ] \(pytorch/pytorch#20572) Add aten mkldnn batchnorm backward operator 💛
 - [ ] \(pytorch/pytorch#20573) Add aten mkldnn zero_ operator:yellow_heart:
 - [ ] \(pytorch/pytorch#20575) Add mkldnn mul operator 💛
Pull Request resolved: pytorch/pytorch#20575

Differential Revision: D15799529

Pulled By: bddppq

fbshipit-source-id: 4887d8ef1a0e316ad9db199b657d9481fc13e486
facebook-github-bot pushed a commit that referenced this pull request Jun 14, 2019
@XiaobingSuper @facebook-github-bot
Add aten mkldnn zero_ operator (#20573)
c06ccbe 
Summary:
### mkldnn backward ops list:
 - [ ] \(#20567) Add aten mkldnn conv2d backward operator 💛
 - [ ] \(#20570) Add aten mkldnn backward ops: relu, linear and reshape 💛
 - [ ] \(#20571) Add aten mkldnn backward ops: max_pool2d, avg_pool2d and adaptive_avg_poo2d 💛
 - [ ] \(#20572) Add aten mkldnn batchnorm backward operator 💛
 - [ ] \(#20573) Add aten mkldnn zero_ operator:yellow_heart:
 - [ ] \(#20575) Add mkldnn mul operator 💚
Pull Request resolved: #20573

Differential Revision: D15820477

Pulled By: bddppq

fbshipit-source-id: 35d95f5b4e013c8db1911f52148550a2e40a2e68
zdevito pushed a commit to zdevito/ATen that referenced this pull request Jun 14, 2019
@XiaobingSuper @facebook-github-bot
Add aten mkldnn zero_ operator (#20573)
b3c29fc 
Summary:
### mkldnn backward ops list:
 - [ ] \(pytorch/pytorch#20567) Add aten mkldnn conv2d backward operator 💛
 - [ ] \(pytorch/pytorch#20570) Add aten mkldnn backward ops: relu, linear and reshape 💛
 - [ ] \(pytorch/pytorch#20571) Add aten mkldnn backward ops: max_pool2d, avg_pool2d and adaptive_avg_poo2d 💛
 - [ ] \(pytorch/pytorch#20572) Add aten mkldnn batchnorm backward operator 💛
 - [ ] \(pytorch/pytorch#20573) Add aten mkldnn zero_ operator:yellow_heart:
 - [ ] \(pytorch/pytorch#20575) Add mkldnn mul operator 💚
Pull Request resolved: pytorch/pytorch#20573

Differential Revision: D15820477

Pulled By: bddppq

fbshipit-source-id: 35d95f5b4e013c8db1911f52148550a2e40a2e68
@Jianhui-Li
Copy link

Jianhui-Li commented Jun 19, 2019

@gottbrath @bddppq @dzhulgakov @wesolwsk @VitalyFedyunin @ezyang This batch of PRs enabled the MKL-DNN for the training path and is expected to improve training models like resnext101 by 2x on CPU with AVX512 (also benefit AVX2). With more DL accelerating instructions like bfp16 in CPU roadmap, we think it is important to accelerate the Pytorch training path as we do for the inference path so the community get the benefits.

@dzhulgakov
Copy link
Collaborator

dzhulgakov commented Jul 15, 2019

Hey, @XiaobingSuper - is it possible to rebase these PRs on top of current master and make sure the CI passes?

@XiaobingSuper XiaobingSuper force-pushed the mkldnn_pool_bwd branch 4 times, most recently from 92aa6eb to 93c3595 Compare July 18, 2019 05:11
@dariogarcia
Copy link

dariogarcia commented Sep 6, 2019

My research group is really interested in this. We are using PyTorch on CPUs because our research has extreme memory requirements, and TensorFlow has a limited tensor size whereas PyTorch does not.

@gbaydin
Copy link

gbaydin commented Sep 6, 2019

Hi, our group is interested in this PR to improve CPU training in our large-scale probabilistic programming work.

@wbhimji
Copy link

wbhimji commented Oct 11, 2019

I'd like to also support that we need these optimizations for the PyTorch we run in support of large-scale science at the NERSC supercomputing center. Currently we blocking on using more recent PyTorch 1.2 and 1.3 for some projects because of this. Thanks!

@XiaobingSuper
Copy link
Collaborator Author

XiaobingSuper commented Oct 12, 2019

@dariogarcia, @gbaydin, @wbhimji, thank you for your interest in our work. I will rebase the code after get the reponse from Facebook team.

@XiaobingSuper
Copy link
Collaborator Author

XiaobingSuper commented Oct 12, 2019

@VitalyFedyunin

@DavidRuhe
Copy link

DavidRuhe commented Oct 12, 2019

We at SURF (national super-computing facilities in the Netherlands) are also really interested in better MKL support in PyTorch. In our case specifically 3D operations.

@XiaobingSuper
Copy link
Collaborator Author

XiaobingSuper commented Oct 25, 2019

@VitalyFedyunin

VitalyFedyunin
VitalyFedyunin reviewed Mar 31, 2020
View reviewed changes
Copy link
Contributor

@VitalyFedyunin VitalyFedyunin left a comment

Choose a reason for hiding this comment

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

Overall looks good, can you please rebase for final tests. Also what is the plan to start renaming functions to onednn?

@XiaobingSuper
Copy link
Collaborator Author

XiaobingSuper commented Apr 1, 2020

Overall looks good, can you please rebase for final tests. Also what is the plan to start renaming functions to onednn?

@VitalyFedyunin , the renaming function will be ok at next step which will upgrade DNNL to v1.4, so I will rebase those PRs which stiil using the old name, thanks！

@XiaobingSuper XiaobingSuper mentioned this pull request Apr 7, 2020
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@gchanan gchanan Awaiting requested review from gchanan

@bddppq bddppq Awaiting requested review from bddppq

@dzhulgakov dzhulgakov Awaiting requested review from dzhulgakov

@apaszke apaszke Awaiting requested review from apaszke

1 more reviewer

@VitalyFedyunin VitalyFedyunin VitalyFedyunin left review comments

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

module: autograd Related to torch.autograd, and the autograd engine in general module: internals Related to internal abstractions in c10 and ATen module: mkldnn Related to Intel IDEEP or oneDNN (a.k.a. mkldnn) integration open source triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.