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

add interpolate_like for cpu #10544

Open
wants to merge 18 commits into
base: master
Choose a base branch
from
Open

add interpolate_like for cpu #10544

wants to merge 18 commits into from

Conversation

woaixiaoxiao
Copy link
Contributor

@woaixiaoxiao woaixiaoxiao commented Jul 9, 2024 •
edited
Loading

在npu上进行SD的推理时发现用到了interpolate_like算子，但在开源版本中还没提供该算子的cpu实现。

因此该pr将闭源版本的interpolate_like算子的正向计算过程补充到开源版本。

  • 在该分支下编译 oneflow，就可以在 cpu 和 npu 端调用 interpolate_like 算子了

@woaixiaoxiao woaixiaoxiao requested a review from hjchen2 as a code owner July 9, 2024 08:23
@github-actions GitHub Actions
Copy link
Contributor

github-actions bot commented Jul 9, 2024

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 GitHub Actions
Copy link
Contributor

github-actions bot commented Jul 9, 2024

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.

Flowingsun007
Flowingsun007 reviewed Jul 10, 2024
View reviewed changes
python/oneflow/nn/modules/interpolate_like.py
def interpolate_like(
input, like, mode="nearest", align_corners=None,
):
"""The interface is consistent with PyTorch.
Copy link
Contributor

Choose a reason for hiding this comment

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

参考interpolate，需要在docs/source/nn.functional.rst 里加一下interpolate_like

Copy link
Contributor Author

Choose a reason for hiding this comment

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

已补充

@woaixiaoxiao woaixiaoxiao requested a review from doombeaker as a code owner July 11, 2024 01:58
@ShawnXuan ShawnXuan requested a review from oneflow-ci-bot July 11, 2024 02:05
@Flowingsun007 Flowingsun007 enabled auto-merge (squash) July 15, 2024 04:00
@github-actions GitHub Actions
Copy link
Contributor

View latest API docs preview at: https://oneflow-staging.oss-cn-beijing.aliyuncs.com/docs/Oneflow-Inc/oneflow/pr/10544/

@github-actions GitHub Actions
Copy link
Contributor

Speed stats:

@ShawnXuan ShawnXuan self-requested a review August 2, 2024 09:47
@github-actions GitHub Actions
Copy link
Contributor

github-actions bot commented Aug 2, 2024

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

@github-actions github-actions bot removed the automerge label Aug 2, 2024
@github-actions GitHub Actions
Copy link
Contributor

github-actions bot commented Aug 4, 2024

View latest API docs preview at: https://oneflow-staging.oss-cn-beijing.aliyuncs.com/docs/Oneflow-Inc/oneflow/pr/10544/

@github-actions GitHub Actions
Copy link
Contributor

github-actions bot commented Aug 4, 2024

Speed stats: 
GPU Name: NVIDIA GeForce RTX 3080 Ti 

❌ OneFlow resnet50 time: 43.3ms (= 4326.7ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 57.4ms (= 5743.8ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.33 (= 57.4ms / 43.3ms)

OneFlow resnet50 time: 26.4ms (= 2635.4ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 37.3ms (= 3729.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.42 (= 37.3ms / 26.4ms)

OneFlow resnet50 time: 18.3ms (= 3664.8ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 35.0ms (= 7000.3ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.91 (= 35.0ms / 18.3ms)

OneFlow resnet50 time: 17.2ms (= 3439.3ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 31.5ms (= 6304.8ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.83 (= 31.5ms / 17.2ms)

OneFlow resnet50 time: 17.0ms (= 3398.1ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 29.2ms (= 5844.5ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.72 (= 29.2ms / 17.0ms)

OneFlow swin dataloader time: 0.200s (= 40.028s / 200, num_workers=1)
PyTorch swin dataloader time: 0.129s (= 25.747s / 200, num_workers=1)
Relative speed: 0.643 (= 0.129s / 0.200s)

OneFlow swin dataloader time: 0.054s (= 10.721s / 200, num_workers=4)
PyTorch swin dataloader time: 0.033s (= 6.686s / 200, num_workers=4)
Relative speed: 0.624 (= 0.033s / 0.054s)

OneFlow swin dataloader time: 0.030s (= 6.047s / 200, num_workers=8)
PyTorch swin dataloader time: 0.017s (= 3.340s / 200, num_workers=8)
Relative speed: 0.552 (= 0.017s / 0.030s)

❌ OneFlow resnet50 time: 49.1ms (= 4905.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 64.0ms (= 6400.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 64.0ms / 49.1ms)

OneFlow resnet50 time: 36.9ms (= 3693.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 46.1ms (= 4609.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.25 (= 46.1ms / 36.9ms)

OneFlow resnet50 time: 27.7ms (= 5540.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 40.8ms (= 8152.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.47 (= 40.8ms / 27.7ms)

OneFlow resnet50 time: 25.3ms (= 5051.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 38.6ms (= 7722.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.53 (= 38.6ms / 25.3ms)

OneFlow resnet50 time: 24.8ms (= 4969.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 35.8ms (= 7169.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.44 (= 35.8ms / 24.8ms)

@github-actions GitHub Actions
Copy link
Contributor

github-actions bot commented Aug 8, 2024

Speed stats: 
GPU Name: NVIDIA GeForce RTX 3080 Ti 

❌ OneFlow resnet50 time: 43.3ms (= 4327.0ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 57.3ms (= 5733.0ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.32 (= 57.3ms / 43.3ms)

OneFlow resnet50 time: 26.1ms (= 2605.7ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 38.1ms (= 3806.4ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.46 (= 38.1ms / 26.1ms)

OneFlow resnet50 time: 17.6ms (= 3526.9ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 37.1ms (= 7429.9ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 2.11 (= 37.1ms / 17.6ms)

OneFlow resnet50 time: 16.9ms (= 3384.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 31.6ms (= 6314.3ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.87 (= 31.6ms / 16.9ms)

OneFlow resnet50 time: 17.3ms (= 3463.1ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 31.5ms (= 6291.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.82 (= 31.5ms / 17.3ms)

OneFlow swin dataloader time: 0.200s (= 40.004s / 200, num_workers=1)
PyTorch swin dataloader time: 0.130s (= 25.939s / 200, num_workers=1)
Relative speed: 0.648 (= 0.130s / 0.200s)

OneFlow swin dataloader time: 0.053s (= 10.569s / 200, num_workers=4)
PyTorch swin dataloader time: 0.032s (= 6.431s / 200, num_workers=4)
Relative speed: 0.609 (= 0.032s / 0.053s)

OneFlow swin dataloader time: 0.030s (= 5.965s / 200, num_workers=8)
PyTorch swin dataloader time: 0.017s (= 3.382s / 200, num_workers=8)
Relative speed: 0.567 (= 0.017s / 0.030s)

❌ OneFlow resnet50 time: 49.3ms (= 4928.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 64.0ms (= 6397.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 64.0ms / 49.3ms)

OneFlow resnet50 time: 36.1ms (= 3608.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 45.6ms (= 4559.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.26 (= 45.6ms / 36.1ms)

OneFlow resnet50 time: 27.8ms (= 5550.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 39.8ms (= 7964.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.43 (= 39.8ms / 27.8ms)

OneFlow resnet50 time: 25.2ms (= 5035.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 40.4ms (= 8070.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.60 (= 40.4ms / 25.2ms)

OneFlow resnet50 time: 24.7ms (= 4945.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 35.8ms (= 7158.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.45 (= 35.8ms / 24.7ms)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@Flowingsun007 Flowingsun007 Flowingsun007 approved these changes

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

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

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

@ShawnXuan ShawnXuan Awaiting requested review from ShawnXuan

Assignees
No one assigned
Labels
enhancement op
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@woaixiaoxiao @ShawnXuan @Flowingsun007 @oneflow-ci-bot