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

[CodeGen]Support AppendOneFlowStream Pass For MLIR #10149

Merged
merged 29 commits into from Apr 23, 2023
Merged

[CodeGen]Support AppendOneFlowStream Pass For MLIR #10149

merged 29 commits into from Apr 23, 2023

Conversation

howin98
Copy link
Contributor

@howin98 howin98 commented Apr 17, 2023

No description provided.

howin98 and others added 5 commits April 17, 2023 03:50
@howin98
trim
61508a2
@howin98
fix
d78d84d
@howin98 @oneflow-ci-bot
@daquexian @mergify
speed up developing ofstream in mlir codegen (#10152)
aa22da4 
Signed-off-by: daquexian <daquexian566@gmail.com>
Co-authored-by: oneflow-ci-bot <ci-bot@oneflow.org>
Co-authored-by: daquexian <daquexian566@gmail.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
@howin98
[CodeGen]Support MgpuToOneFlowStream Pass For MLIR (#10151)
1d14a51
@oneflow-ci-bot
auto format by CI
7ad3de2
@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: Build cpu. PR label automerge has been removed

@howin98 howin98 enabled auto-merge (squash) April 20, 2023 23:38
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.4ms (= 14143.5ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 147.3ms (= 14727.5ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.04 (= 147.3ms / 141.4ms)

OneFlow resnet50 time: 82.7ms (= 8270.3ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 94.6ms (= 9462.6ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.14 (= 94.6ms / 82.7ms)

OneFlow resnet50 time: 51.6ms (= 10317.6ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 71.9ms (= 14387.6ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.39 (= 71.9ms / 51.6ms)

OneFlow resnet50 time: 34.0ms (= 6806.5ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 65.2ms (= 13030.1ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.91 (= 65.2ms / 34.0ms)

OneFlow resnet50 time: 26.2ms (= 5241.6ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 63.5ms (= 12708.7ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.42 (= 63.5ms / 26.2ms)

OneFlow swin dataloader time: 0.258s (= 51.622s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.113s / 200, num_workers=1)
Relative speed: 0.583 (= 0.151s / 0.258s)

OneFlow swin dataloader time: 0.071s (= 14.204s / 200, num_workers=4)
PyTorch swin dataloader time: 0.042s (= 8.430s / 200, num_workers=4)
Relative speed: 0.593 (= 0.042s / 0.071s)

OneFlow swin dataloader time: 0.041s (= 8.288s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.672s / 200, num_workers=8)
Relative speed: 0.564 (= 0.023s / 0.041s)

❌ OneFlow resnet50 time: 153.8ms (= 15378.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 165.2ms (= 16518.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 165.2ms / 153.8ms)

OneFlow resnet50 time: 94.4ms (= 9435.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.6ms (= 10459.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 104.6ms / 94.4ms)

OneFlow resnet50 time: 61.6ms (= 12317.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.9ms (= 15776.1ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 78.9ms / 61.6ms)

OneFlow resnet50 time: 44.0ms (= 8796.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.8ms (= 14164.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.61 (= 70.8ms / 44.0ms)

OneFlow resnet50 time: 38.5ms (= 7708.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.0ms (= 14395.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.87 (= 72.0ms / 38.5ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10149/

@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: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.4ms (= 14144.5ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 146.8ms (= 14676.6ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.04 (= 146.8ms / 141.4ms)

OneFlow resnet50 time: 83.1ms (= 8305.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 93.8ms (= 9380.7ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.13 (= 93.8ms / 83.1ms)

OneFlow resnet50 time: 52.4ms (= 10470.5ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 71.2ms (= 14249.9ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.36 (= 71.2ms / 52.4ms)

OneFlow resnet50 time: 33.9ms (= 6773.3ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 66.8ms (= 13364.5ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.97 (= 66.8ms / 33.9ms)

OneFlow resnet50 time: 26.6ms (= 5317.4ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 63.1ms (= 12613.2ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.37 (= 63.1ms / 26.6ms)

OneFlow swin dataloader time: 0.247s (= 49.407s / 200, num_workers=1)
PyTorch swin dataloader time: 0.148s (= 29.612s / 200, num_workers=1)
Relative speed: 0.599 (= 0.148s / 0.247s)

OneFlow swin dataloader time: 0.073s (= 14.662s / 200, num_workers=4)
PyTorch swin dataloader time: 0.043s (= 8.578s / 200, num_workers=4)
Relative speed: 0.585 (= 0.043s / 0.073s)

OneFlow swin dataloader time: 0.042s (= 8.399s / 200, num_workers=8)
PyTorch swin dataloader time: 0.024s (= 4.704s / 200, num_workers=8)
Relative speed: 0.560 (= 0.024s / 0.042s)

❌ OneFlow resnet50 time: 165.1ms (= 16505.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 174.6ms (= 17457.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.06 (= 174.6ms / 165.1ms)

OneFlow resnet50 time: 104.0ms (= 10395.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 112.9ms (= 11285.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.09 (= 112.9ms / 104.0ms)

OneFlow resnet50 time: 72.1ms (= 14418.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 95.8ms (= 19159.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.33 (= 95.8ms / 72.1ms)

OneFlow resnet50 time: 59.3ms (= 11850.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.6ms (= 15122.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 75.6ms / 59.3ms)

OneFlow resnet50 time: 52.8ms (= 10558.7ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.9ms (= 15382.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.46 (= 76.9ms / 52.8ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10149/

@github-actions
Copy link
Contributor

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

@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: 141.3ms (= 14126.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 146.3ms (= 14634.0ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.04 (= 146.3ms / 141.3ms)

OneFlow resnet50 time: 81.4ms (= 8140.4ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 91.3ms (= 9129.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.12 (= 91.3ms / 81.4ms)

OneFlow resnet50 time: 51.3ms (= 10256.4ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 71.4ms (= 14272.7ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.39 (= 71.4ms / 51.3ms)

OneFlow resnet50 time: 34.2ms (= 6841.2ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 63.7ms (= 12743.3ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.86 (= 63.7ms / 34.2ms)

OneFlow resnet50 time: 26.4ms (= 5270.4ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 64.5ms (= 12892.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.45 (= 64.5ms / 26.4ms)

OneFlow swin dataloader time: 0.245s (= 49.031s / 200, num_workers=1)
PyTorch swin dataloader time: 0.152s (= 30.378s / 200, num_workers=1)
Relative speed: 0.620 (= 0.152s / 0.245s)

OneFlow swin dataloader time: 0.071s (= 14.207s / 200, num_workers=4)
PyTorch swin dataloader time: 0.040s (= 7.971s / 200, num_workers=4)
Relative speed: 0.561 (= 0.040s / 0.071s)

OneFlow swin dataloader time: 0.042s (= 8.499s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.457s / 200, num_workers=8)
Relative speed: 0.524 (= 0.022s / 0.042s)

❌ OneFlow resnet50 time: 152.8ms (= 15284.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 165.7ms (= 16573.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.08 (= 165.7ms / 152.8ms)

OneFlow resnet50 time: 93.2ms (= 9317.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.6ms (= 10359.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 103.6ms / 93.2ms)

OneFlow resnet50 time: 61.6ms (= 12323.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 80.1ms (= 16020.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 80.1ms / 61.6ms)

OneFlow resnet50 time: 43.5ms (= 8690.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.3ms (= 15451.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.78 (= 77.3ms / 43.5ms)

OneFlow resnet50 time: 37.3ms (= 7453.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.3ms (= 14066.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.89 (= 70.3ms / 37.3ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10149/

@howin98 howin98 merged commit 14e2cb4 into master Apr 23, 2023
22 checks passed
@howin98 howin98 deleted the support-append-ofstream-pass branch April 23, 2023 08:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@mosout mosout mosout approved these changes

@BBuf BBuf BBuf approved these changes

@daquexian daquexian Awaiting requested review from daquexian

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

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

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

Assignees
No one assigned
Labels
automerge enhancement ir need-all-tests-even-fail
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
4 participants
@howin98 @mosout @BBuf @oneflow-ci-bot