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

support copy_ on different devices and dtypes #9888

Merged
merged 16 commits into from
Feb 28, 2023
Merged

support copy_ on different devices and dtypes #9888

merged 16 commits into from
Feb 28, 2023

Conversation

daquexian
Copy link
Contributor

支持 copy_ 的输入和输出的 device 和 dtype 都不同的情况

@daquexian
support copy_ on different devices and dtypes
5d09ab3 
Signed-off-by: daquexian <daquexian566@gmail.com>
@daquexian daquexian marked this pull request as ready for review February 22, 2023 10:58
@daquexian daquexian requested a review from BBuf as a code owner February 22, 2023 10:58
@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.

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats:

@github-actions
Copy link
Contributor

Speed stats:

@github-actions
Copy link
Contributor

CI failed when running job: cpu-module. PR label automerge has been removed

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats:

@daquexian daquexian requested review from oneflow-ci-bot and removed request for oneflow-ci-bot February 23, 2023 03:15
@github-actions
Copy link
Contributor

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

daquexian and others added 3 commits February 27, 2023 13:04
@daquexian
Merge branch 'master' into inplace_copy_diff_dev_dtype
88394b1 
Signed-off-by: daquexian <daquexian566@gmail.com>
@daquexian
fix tests
99954e5 
Signed-off-by: daquexian <daquexian566@gmail.com>
@oneflow-ci-bot
auto format by CI
72a2560
@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.

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats:

@daquexian
fix tests
31b5932 
Signed-off-by: daquexian <daquexian566@gmail.com>
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.1ms (= 14110.7ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 144.7ms (= 14472.3ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.03 (= 144.7ms / 141.1ms)

OneFlow resnet50 time: 82.4ms (= 8237.8ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 87.9ms (= 8792.6ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 87.9ms / 82.4ms)

OneFlow resnet50 time: 50.9ms (= 10186.5ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 54.4ms (= 10871.3ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.07 (= 54.4ms / 50.9ms)

OneFlow resnet50 time: 34.6ms (= 6921.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.8ms (= 8763.3ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.27 (= 43.8ms / 34.6ms)

OneFlow resnet50 time: 26.9ms (= 5372.2ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 42.2ms (= 8442.2ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.57 (= 42.2ms / 26.9ms)

OneFlow swin dataloader time: 0.242s (= 48.356s / 200, num_workers=1)
PyTorch swin dataloader time: 0.148s (= 29.599s / 200, num_workers=1)
Relative speed: 0.612 (= 0.148s / 0.242s)

OneFlow swin dataloader time: 0.070s (= 13.990s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.239s / 200, num_workers=4)
Relative speed: 0.589 (= 0.041s / 0.070s)

OneFlow swin dataloader time: 0.041s (= 8.152s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.541s / 200, num_workers=8)
Relative speed: 0.557 (= 0.023s / 0.041s)

❌ OneFlow resnet50 time: 153.0ms (= 15299.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 164.4ms (= 16435.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 164.4ms / 153.0ms)

OneFlow resnet50 time: 93.1ms (= 9314.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.4ms (= 10444.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 104.4ms / 93.1ms)

OneFlow resnet50 time: 60.4ms (= 12073.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.8ms (= 15968.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.32 (= 79.8ms / 60.4ms)

OneFlow resnet50 time: 42.4ms (= 8473.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.3ms (= 13868.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.64 (= 69.3ms / 42.4ms)

OneFlow resnet50 time: 36.2ms (= 7241.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.2ms (= 13643.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.88 (= 68.2ms / 36.2ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9888/

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.0ms (= 14101.2ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 143.8ms (= 14384.1ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.02 (= 143.8ms / 141.0ms)

OneFlow resnet50 time: 80.9ms (= 8088.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 85.0ms (= 8499.0ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.05 (= 85.0ms / 80.9ms)

OneFlow resnet50 time: 50.1ms (= 10023.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.7ms (= 11744.4ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.17 (= 58.7ms / 50.1ms)

OneFlow resnet50 time: 33.1ms (= 6614.9ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 44.0ms (= 8790.3ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.33 (= 44.0ms / 33.1ms)

OneFlow resnet50 time: 25.8ms (= 5163.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 42.0ms (= 8410.0ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.63 (= 42.0ms / 25.8ms)

OneFlow swin dataloader time: 0.237s (= 47.385s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.100s / 200, num_workers=1)
Relative speed: 0.635 (= 0.151s / 0.237s)

OneFlow swin dataloader time: 0.068s (= 13.693s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.184s / 200, num_workers=4)
Relative speed: 0.598 (= 0.041s / 0.068s)

OneFlow swin dataloader time: 0.043s (= 8.601s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.424s / 200, num_workers=8)
Relative speed: 0.514 (= 0.022s / 0.043s)

❌ OneFlow resnet50 time: 153.0ms (= 15298.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 160.6ms (= 16057.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.05 (= 160.6ms / 153.0ms)

OneFlow resnet50 time: 92.4ms (= 9243.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.8ms (= 10284.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 102.8ms / 92.4ms)

OneFlow resnet50 time: 59.8ms (= 11962.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.3ms (= 15664.1ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 78.3ms / 59.8ms)

OneFlow resnet50 time: 42.1ms (= 8419.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.7ms (= 14533.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.73 (= 72.7ms / 42.1ms)

OneFlow resnet50 time: 35.6ms (= 7119.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.7ms (= 14338.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 2.01 (= 71.7ms / 35.6ms)

@github-actions
Copy link
Contributor

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

@daquexian
fix tests
2ed2039 
Signed-off-by: daquexian <daquexian566@gmail.com>
@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats:

@daquexian daquexian requested review from oneflow-ci-bot and removed request for oneflow-ci-bot February 28, 2023 13:15
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.2ms (= 14115.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 141.5ms (= 14146.6ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.00 (= 141.5ms / 141.2ms)

OneFlow resnet50 time: 81.6ms (= 8163.2ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 85.7ms (= 8566.3ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.05 (= 85.7ms / 81.6ms)

OneFlow resnet50 time: 50.4ms (= 10089.8ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 61.2ms (= 12237.8ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.21 (= 61.2ms / 50.4ms)

OneFlow resnet50 time: 33.4ms (= 6687.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 43.6ms (= 8717.8ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.30 (= 43.6ms / 33.4ms)

OneFlow resnet50 time: 25.5ms (= 5101.2ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.5ms (= 7497.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.47 (= 37.5ms / 25.5ms)

OneFlow swin dataloader time: 0.234s (= 46.829s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.749s / 200, num_workers=1)
Relative speed: 0.635 (= 0.149s / 0.234s)

OneFlow swin dataloader time: 0.069s (= 13.746s / 200, num_workers=4)
PyTorch swin dataloader time: 0.043s (= 8.530s / 200, num_workers=4)
Relative speed: 0.621 (= 0.043s / 0.069s)

OneFlow swin dataloader time: 0.043s (= 8.513s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.541s / 200, num_workers=8)
Relative speed: 0.533 (= 0.023s / 0.043s)

❌ OneFlow resnet50 time: 152.8ms (= 15279.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.1ms (= 16205.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.06 (= 162.1ms / 152.8ms)

OneFlow resnet50 time: 92.4ms (= 9240.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.5ms (= 10353.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 103.5ms / 92.4ms)

OneFlow resnet50 time: 60.2ms (= 12033.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.0ms (= 15801.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 79.0ms / 60.2ms)

OneFlow resnet50 time: 42.6ms (= 8512.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.3ms (= 14269.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.68 (= 71.3ms / 42.6ms)

OneFlow resnet50 time: 35.8ms (= 7165.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.1ms (= 14220.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.98 (= 71.1ms / 35.8ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9888/

@mergify mergify bot merged commit d6038c6 into master Feb 28, 2023
@mergify mergify bot deleted the inplace_copy_diff_dev_dtype branch February 28, 2023 14:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@CPFLAME CPFLAME CPFLAME approved these changes

@mosout mosout mosout approved these changes

@BBuf BBuf Awaiting requested review from BBuf

@oneflow-ci-bot oneflow-ci-bot Awaiting requested review from oneflow-ci-bot

Assignees
No one assigned
Labels
automerge bug python
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@daquexian @CPFLAME @mosout @oneflow-ci-bot