Skip to content
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

Dev pure scalar div #7483

Merged
merged 28 commits into from
Feb 14, 2022
Merged

Dev pure scalar div #7483

merged 28 commits into from
Feb 14, 2022

Conversation

simonJJJ
Copy link
Contributor

Add pure scalar div for the same floating-point accuracy with torch.

simonJJJ added 19 commits October 4, 2021 22:51
@simonJJJ
fix typo
38b69b3
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
8ee6465
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
a5af37b
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
da3502e
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
2681ad3
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
bf9ec14
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
70bb1e9
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
2b53b9a
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
fbf6daa
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
6d9d0f9
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
de7975e
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
0885fd9
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
eb26686
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
80fe5df
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
2a7fbfe
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
9c396c8
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
f351cd2
@simonJJJ
Merge branch 'master' of https://github.com/Oneflow-Inc/oneflow
4f85ecd
@simonJJJ
add pure scalar div
6c21940
@simonJJJ simonJJJ requested review from BBuf and MARD1NO February 11, 2022 11:51
@github-actions
Copy link
Contributor

Static analysis with clang failed. PR label automerge has been removed

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.9ms (= 12886.8ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 139.9ms (= 13985.1ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.09 (= 139.9ms / 128.9ms)

✔️ OneFlow resnet50 time: 77.1ms (= 7708.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 85.2ms (= 8520.8ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.11 (= 85.2ms / 77.1ms)

OneFlow resnet50 time: 51.6ms (= 10312.2ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 56.6ms (= 11321.5ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.10 (= 56.6ms / 51.6ms)

OneFlow resnet50 time: 44.5ms (= 8905.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 48.5ms (= 9709.2ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.09 (= 48.5ms / 44.5ms)

OneFlow resnet50 time: 41.1ms (= 8227.5ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.4ms (= 7674.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.93 (= 38.4ms / 41.1ms)

✔️ OneFlow resnet50 time: 143.0ms (= 14295.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 159.4ms (= 15936.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 159.4ms / 143.0ms)

OneFlow resnet50 time: 87.8ms (= 8779.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 105.3ms (= 10526.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 105.3ms / 87.8ms)

OneFlow resnet50 time: 61.2ms (= 12249.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 74.7ms (= 14941.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 74.7ms / 61.2ms)

OneFlow resnet50 time: 55.4ms (= 11085.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 63.9ms (= 12774.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 63.9ms / 55.4ms)

OneFlow resnet50 time: 55.5ms (= 11098.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 63.8ms (= 12765.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 63.8ms / 55.5ms)

@oneflow-ci-bot oneflow-ci-bot removed their request for review February 14, 2022 06:35
@github-actions
Copy link
Contributor

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

@simonJJJ simonJJJ requested review from oneflow-ci-bot and removed request for oneflow-ci-bot February 14, 2022 07:50
@oneflow-ci-bot oneflow-ci-bot requested review from oneflow-ci-bot and removed request for oneflow-ci-bot February 14, 2022 07:56
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 129.1ms (= 12914.3ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 141.3ms (= 14131.4ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.09 (= 141.3ms / 129.1ms)

✔️ OneFlow resnet50 time: 78.3ms (= 7830.0ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 86.7ms (= 8670.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.11 (= 86.7ms / 78.3ms)

OneFlow resnet50 time: 53.1ms (= 10610.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.0ms (= 11796.9ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.11 (= 59.0ms / 53.1ms)

OneFlow resnet50 time: 44.7ms (= 8937.1ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.3ms (= 8667.8ms / 200, input_shape=[2, 3, 224, 224])
❌ Relative speed: 0.97 (= 43.3ms / 44.7ms)

OneFlow resnet50 time: 39.6ms (= 7917.6ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.7ms (= 7537.3ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.95 (= 37.7ms / 39.6ms)

✔️ OneFlow resnet50 time: 142.6ms (= 14257.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.9ms (= 16287.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.14 (= 162.9ms / 142.6ms)

OneFlow resnet50 time: 87.4ms (= 8739.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.5ms (= 10349.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 103.5ms / 87.4ms)

OneFlow resnet50 time: 61.8ms (= 12352.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.4ms (= 15280.1ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.24 (= 76.4ms / 61.8ms)

OneFlow resnet50 time: 52.6ms (= 10513.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.0ms (= 13808.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 69.0ms / 52.6ms)

OneFlow resnet50 time: 50.2ms (= 10040.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 62.0ms (= 12405.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.24 (= 62.0ms / 50.2ms)

@github-actions
Copy link
Contributor

CI failed when running job: cuda-speed-test. PR label automerge has been removed

@oneflow-ci-bot oneflow-ci-bot removed their request for review February 14, 2022 09:03
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 129.0ms (= 12898.2ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 140.8ms (= 14076.4ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.09 (= 140.8ms / 129.0ms)

✔️ OneFlow resnet50 time: 78.0ms (= 7802.0ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 87.5ms (= 8745.0ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.12 (= 87.5ms / 78.0ms)

OneFlow resnet50 time: 53.5ms (= 10707.5ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.9ms (= 11789.3ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.10 (= 58.9ms / 53.5ms)

OneFlow resnet50 time: 44.6ms (= 8925.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 49.6ms (= 9911.9ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.11 (= 49.6ms / 44.6ms)

OneFlow resnet50 time: 39.9ms (= 7979.8ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 43.4ms (= 8678.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.09 (= 43.4ms / 39.9ms)

✔️ OneFlow resnet50 time: 142.1ms (= 14212.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 158.4ms (= 15838.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 158.4ms / 142.1ms)

OneFlow resnet50 time: 89.6ms (= 8957.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.7ms (= 10267.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 102.7ms / 89.6ms)

OneFlow resnet50 time: 62.4ms (= 12479.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 84.3ms (= 16866.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 84.3ms / 62.4ms)

OneFlow resnet50 time: 51.7ms (= 10330.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.7ms (= 13740.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.33 (= 68.7ms / 51.7ms)

OneFlow resnet50 time: 53.8ms (= 10769.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 65.6ms (= 13110.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 65.6ms / 53.8ms)

@oneflow-ci-bot oneflow-ci-bot removed their request for review February 14, 2022 10:04
@oneflow-ci-bot oneflow-ci-bot merged commit 4a4f5d1 into master Feb 14, 2022
@oneflow-ci-bot oneflow-ci-bot deleted the dev_pure_scalar_div branch February 14, 2022 10:04
marigoold pushed a commit that referenced this pull request Mar 15, 2022
@simonJJJ @oneflow-ci-bot @marigoold
Dev pure scalar div (#7483)
03562ea 
* fix typo

* add pure scalar div

* fix static check

* fix vectorat

* auto format by CI

Co-authored-by: oneflow-ci-bot <69100618+oneflow-ci-bot@users.noreply.github.com>
Co-authored-by: oneflow-ci-bot <ci-bot@oneflow.org>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@MARD1NO MARD1NO MARD1NO approved these changes

@BBuf BBuf BBuf approved these changes

@hjchen2 hjchen2 Awaiting requested review from hjchen2 hjchen2 is a code owner

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

@guo-ran guo-ran Awaiting requested review from guo-ran

Assignees
No one assigned
Labels
api automerge eager feature
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@simonJJJ @BBuf @MARD1NO @oneflow-ci-bot