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

refactor_ccl_comm_mgr_register #10100

Merged
merged 17 commits into from
Apr 21, 2023
Merged

refactor_ccl_comm_mgr_register #10100

merged 17 commits into from
Apr 21, 2023

Conversation

clackhan
Copy link
Contributor

No description provided.

@clackhan clackhan requested review from leaves-zwx and hjchen2 April 10, 2023 10:16
@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

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

Speed stats: 
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.5ms (= 14146.4ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 148.4ms (= 14838.6ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.05 (= 148.4ms / 141.5ms)

OneFlow resnet50 time: 83.6ms (= 8355.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 93.7ms (= 9366.7ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.12 (= 93.7ms / 83.6ms)

OneFlow resnet50 time: 51.5ms (= 10299.1ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 72.0ms (= 14403.2ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.40 (= 72.0ms / 51.5ms)

OneFlow resnet50 time: 34.5ms (= 6894.5ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 66.4ms (= 13289.3ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.93 (= 66.4ms / 34.5ms)

OneFlow resnet50 time: 26.7ms (= 5340.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 64.8ms (= 12955.6ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.43 (= 64.8ms / 26.7ms)

OneFlow swin dataloader time: 0.244s (= 48.728s / 200, num_workers=1)
PyTorch swin dataloader time: 0.148s (= 29.680s / 200, num_workers=1)
Relative speed: 0.609 (= 0.148s / 0.244s)

OneFlow swin dataloader time: 0.069s (= 13.791s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.263s / 200, num_workers=4)
Relative speed: 0.599 (= 0.041s / 0.069s)

OneFlow swin dataloader time: 0.042s (= 8.431s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.479s / 200, num_workers=8)
Relative speed: 0.531 (= 0.022s / 0.042s)

❌ OneFlow resnet50 time: 155.4ms (= 15535.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 166.6ms (= 16664.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.07 (= 166.6ms / 155.4ms)

OneFlow resnet50 time: 94.6ms (= 9459.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 103.8ms (= 10379.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.10 (= 103.8ms / 94.6ms)

OneFlow resnet50 time: 62.1ms (= 12426.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 80.0ms (= 15995.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.29 (= 80.0ms / 62.1ms)

OneFlow resnet50 time: 44.0ms (= 8804.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.9ms (= 14170.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.61 (= 70.9ms / 44.0ms)

OneFlow resnet50 time: 40.1ms (= 8024.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.6ms (= 13713.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.71 (= 68.6ms / 40.1ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10100/

@github-actions
Copy link
Contributor

Speed stats:

@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.

@clackhan clackhan changed the title refine_ccl_comm_mgr_create_and_delete refactor_ccl_comm_mgr_register Apr 21, 2023
@hjchen2 hjchen2 enabled auto-merge (squash) April 21, 2023 02:07
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.3ms (= 14134.4ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 147.4ms (= 14743.6ms / 100, input_shape=[16, 3, 224, 224])
❌ Relative speed: 1.04 (= 147.4ms / 141.3ms)

OneFlow resnet50 time: 82.9ms (= 8294.2ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 93.9ms (= 9390.4ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.13 (= 93.9ms / 82.9ms)

OneFlow resnet50 time: 51.7ms (= 10335.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 71.2ms (= 14243.3ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.38 (= 71.2ms / 51.7ms)

OneFlow resnet50 time: 34.8ms (= 6964.5ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 65.3ms (= 13053.9ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.87 (= 65.3ms / 34.8ms)

OneFlow resnet50 time: 27.3ms (= 5450.3ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 56.9ms (= 11380.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 2.09 (= 56.9ms / 27.3ms)

OneFlow swin dataloader time: 0.237s (= 47.434s / 200, num_workers=1)
PyTorch swin dataloader time: 0.149s (= 29.807s / 200, num_workers=1)
Relative speed: 0.628 (= 0.149s / 0.237s)

OneFlow swin dataloader time: 0.068s (= 13.654s / 200, num_workers=4)
PyTorch swin dataloader time: 0.043s (= 8.530s / 200, num_workers=4)
Relative speed: 0.625 (= 0.043s / 0.068s)

OneFlow swin dataloader time: 0.044s (= 8.841s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.473s / 200, num_workers=8)
Relative speed: 0.506 (= 0.022s / 0.044s)

❌ OneFlow resnet50 time: 154.2ms (= 15423.6ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 166.6ms (= 16655.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.08 (= 166.6ms / 154.2ms)

OneFlow resnet50 time: 94.1ms (= 9413.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.9ms (= 10493.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 104.9ms / 94.1ms)

OneFlow resnet50 time: 62.0ms (= 12393.1ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 83.5ms (= 16707.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 83.5ms / 62.0ms)

OneFlow resnet50 time: 43.7ms (= 8749.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 70.0ms (= 13998.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.60 (= 70.0ms / 43.7ms)

OneFlow resnet50 time: 37.8ms (= 7559.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.1ms (= 13427.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.78 (= 67.1ms / 37.8ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10100/

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

❌ OneFlow resnet50 time: 141.6ms (= 14157.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 148.1ms (= 14814.1ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.05 (= 148.1ms / 141.6ms)

OneFlow resnet50 time: 82.9ms (= 8294.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 90.8ms (= 9077.7ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.09 (= 90.8ms / 82.9ms)

OneFlow resnet50 time: 51.7ms (= 10339.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 70.5ms (= 14106.7ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.36 (= 70.5ms / 51.7ms)

OneFlow resnet50 time: 33.9ms (= 6785.2ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 64.6ms (= 12914.2ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.90 (= 64.6ms / 33.9ms)

OneFlow resnet50 time: 26.6ms (= 5319.6ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 34.2ms (= 6831.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.28 (= 34.2ms / 26.6ms)

OneFlow swin dataloader time: 0.241s (= 48.249s / 200, num_workers=1)
PyTorch swin dataloader time: 0.150s (= 29.924s / 200, num_workers=1)
Relative speed: 0.620 (= 0.150s / 0.241s)

OneFlow swin dataloader time: 0.072s (= 14.346s / 200, num_workers=4)
PyTorch swin dataloader time: 0.043s (= 8.506s / 200, num_workers=4)
Relative speed: 0.593 (= 0.043s / 0.072s)

OneFlow swin dataloader time: 0.038s (= 7.614s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.432s / 200, num_workers=8)
Relative speed: 0.582 (= 0.022s / 0.038s)

❌ OneFlow resnet50 time: 154.3ms (= 15428.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 167.4ms (= 16735.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.08 (= 167.4ms / 154.3ms)

OneFlow resnet50 time: 94.4ms (= 9443.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.9ms (= 10487.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.11 (= 104.9ms / 94.4ms)

OneFlow resnet50 time: 61.9ms (= 12373.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 81.9ms (= 16387.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.32 (= 81.9ms / 61.9ms)

OneFlow resnet50 time: 44.7ms (= 8938.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 73.8ms (= 14764.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.65 (= 73.8ms / 44.7ms)

OneFlow resnet50 time: 37.6ms (= 7520.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.1ms (= 13612.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.81 (= 68.1ms / 37.6ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/10100/

@hjchen2 hjchen2 merged commit cd9e4e3 into master Apr 21, 2023
@hjchen2 hjchen2 deleted the refine_ccl_comm_mgr_create_and_delete branch April 21, 2023 13:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@hjchen2 hjchen2 hjchen2 approved these changes

@Flowingsun007 Flowingsun007 Flowingsun007 approved these changes

@leaves-zwx leaves-zwx Awaiting requested review from leaves-zwx

@BBuf BBuf Awaiting requested review from BBuf

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

@daquexian daquexian Awaiting requested review from daquexian

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

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

Successfully merging this pull request may close these issues.
4 participants
@clackhan @hjchen2 @Flowingsun007 @oneflow-ci-bot