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

Alleviate benchmark failure #8043

Merged
merged 7 commits into from
Apr 19, 2022
Merged

Alleviate benchmark failure #8043

merged 7 commits into from
Apr 19, 2022

Conversation

jackalcooper
Copy link
Collaborator

@jackalcooper jackalcooper commented Apr 18, 2022 •
edited
Loading

unknown-threshold 改为 15 个点

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8043/

@jackalcooper jackalcooper enabled auto-merge (squash) April 18, 2022 08:14
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.7ms (= 12869.6ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 142.8ms (= 14280.6ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.11 (= 142.8ms / 128.7ms)

OneFlow resnet50 time: 78.6ms (= 7864.9ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 82.2ms (= 8219.5ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.05 (= 82.2ms / 78.6ms)

OneFlow resnet50 time: 53.1ms (= 10611.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.1ms (= 11627.6ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.10 (= 58.1ms / 53.1ms)

OneFlow resnet50 time: 44.0ms (= 8800.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 49.4ms (= 9880.8ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.12 (= 49.4ms / 44.0ms)

OneFlow resnet50 time: 36.4ms (= 7279.5ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 40.7ms (= 8140.2ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.12 (= 40.7ms / 36.4ms)

OneFlow swin dataloader time: 0.253s (= 50.634s / 200, num_workers=1)
PyTorch swin dataloader time: 0.251s (= 50.232s / 200, num_workers=1)
✔️ Relative speed: 0.992 (= 0.251s / 0.253s)

OneFlow swin dataloader time: 0.069s (= 13.883s / 200, num_workers=4)
PyTorch swin dataloader time: 0.068s (= 13.569s / 200, num_workers=4)
✔️ Relative speed: 0.977 (= 0.068s / 0.069s)

OneFlow swin dataloader time: 0.036s (= 7.298s / 200, num_workers=8)
PyTorch swin dataloader time: 0.037s (= 7.451s / 200, num_workers=8)
✔️ Relative speed: 1.021 (= 0.037s / 0.036s)

✔️ OneFlow resnet50 time: 135.5ms (= 13546.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 156.7ms (= 15673.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 156.7ms / 135.5ms)

OneFlow resnet50 time: 88.6ms (= 8858.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 99.2ms (= 9922.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 99.2ms / 88.6ms)

OneFlow resnet50 time: 59.4ms (= 11878.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.4ms (= 15087.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.27 (= 75.4ms / 59.4ms)

OneFlow resnet50 time: 52.1ms (= 10422.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.1ms (= 13417.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.29 (= 67.1ms / 52.1ms)

OneFlow resnet50 time: 48.4ms (= 9673.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.4ms (= 14482.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.50 (= 72.4ms / 48.4ms)

@oneflow-ci-bot oneflow-ci-bot requested review from oneflow-ci-bot and removed request for oneflow-ci-bot April 18, 2022 08:22
@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.6ms (= 12864.8ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 138.9ms (= 13885.2ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.08 (= 138.9ms / 128.6ms)

OneFlow resnet50 time: 78.4ms (= 7844.5ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 83.8ms (= 8376.1ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 83.8ms / 78.4ms)

OneFlow resnet50 time: 53.1ms (= 10618.9ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.0ms (= 11591.7ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.09 (= 58.0ms / 53.1ms)

OneFlow resnet50 time: 40.5ms (= 8102.5ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 48.2ms (= 9638.2ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.19 (= 48.2ms / 40.5ms)

OneFlow resnet50 time: 39.7ms (= 7936.1ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.6ms (= 7713.5ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.97 (= 38.6ms / 39.7ms)

OneFlow swin dataloader time: 0.248s (= 49.586s / 200, num_workers=1)
PyTorch swin dataloader time: 0.251s (= 50.144s / 200, num_workers=1)
✔️ Relative speed: 1.011 (= 0.251s / 0.248s)

OneFlow swin dataloader time: 0.064s (= 12.837s / 200, num_workers=4)
PyTorch swin dataloader time: 0.079s (= 15.872s / 200, num_workers=4)
✔️ Relative speed: 1.236 (= 0.079s / 0.064s)

OneFlow swin dataloader time: 0.037s (= 7.322s / 200, num_workers=8)
PyTorch swin dataloader time: 0.036s (= 7.264s / 200, num_workers=8)
✔️ Relative speed: 0.992 (= 0.036s / 0.037s)

✔️ OneFlow resnet50 time: 135.5ms (= 13550.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 155.5ms (= 15548.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 155.5ms / 135.5ms)

OneFlow resnet50 time: 88.0ms (= 8802.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 106.0ms (= 10598.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 106.0ms / 88.0ms)

OneFlow resnet50 time: 60.8ms (= 12158.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.2ms (= 15232.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.25 (= 76.2ms / 60.8ms)

OneFlow resnet50 time: 52.2ms (= 10436.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.2ms (= 13437.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.29 (= 67.2ms / 52.2ms)

OneFlow resnet50 time: 47.5ms (= 9502.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 60.7ms (= 12133.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 60.7ms / 47.5ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8043/

@github-actions
Copy link
Contributor

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

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.6ms (= 12864.2ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 138.3ms (= 13833.8ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.08 (= 138.3ms / 128.6ms)

OneFlow resnet50 time: 78.9ms (= 7888.0ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 83.0ms (= 8300.3ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.05 (= 83.0ms / 78.9ms)

OneFlow resnet50 time: 52.4ms (= 10487.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 53.7ms (= 10740.7ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.02 (= 53.7ms / 52.4ms)

OneFlow resnet50 time: 42.0ms (= 8391.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 41.9ms (= 8372.0ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.00 (= 41.9ms / 42.0ms)

OneFlow resnet50 time: 38.6ms (= 7713.0ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.7ms (= 7540.1ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.98 (= 37.7ms / 38.6ms)

OneFlow swin dataloader time: 0.253s (= 50.610s / 200, num_workers=1)
PyTorch swin dataloader time: 0.251s (= 50.126s / 200, num_workers=1)
✔️ Relative speed: 0.990 (= 0.251s / 0.253s)

OneFlow swin dataloader time: 0.068s (= 13.554s / 200, num_workers=4)
PyTorch swin dataloader time: 0.067s (= 13.418s / 200, num_workers=4)
✔️ Relative speed: 0.990 (= 0.067s / 0.068s)

OneFlow swin dataloader time: 0.036s (= 7.278s / 200, num_workers=8)
PyTorch swin dataloader time: 0.038s (= 7.661s / 200, num_workers=8)
✔️ Relative speed: 1.053 (= 0.038s / 0.036s)

✔️ OneFlow resnet50 time: 135.5ms (= 13550.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 156.2ms (= 15623.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 156.2ms / 135.5ms)

OneFlow resnet50 time: 87.6ms (= 8755.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 98.7ms (= 9868.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.13 (= 98.7ms / 87.6ms)

OneFlow resnet50 time: 61.3ms (= 12265.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.2ms (= 15239.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.24 (= 76.2ms / 61.3ms)

OneFlow resnet50 time: 50.9ms (= 10177.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.5ms (= 15299.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.50 (= 76.5ms / 50.9ms)

OneFlow resnet50 time: 48.8ms (= 9768.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 63.6ms (= 12714.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 63.6ms / 48.8ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8043/

@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-benchmark. PR label automerge has been removed

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.2ms (= 12815.7ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 139.2ms (= 13920.7ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.09 (= 139.2ms / 128.2ms)

OneFlow resnet50 time: 78.3ms (= 7826.4ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.0ms (= 8399.4ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 84.0ms / 78.3ms)

OneFlow resnet50 time: 51.8ms (= 10360.5ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 52.1ms (= 10424.4ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.01 (= 52.1ms / 51.8ms)

OneFlow resnet50 time: 43.3ms (= 8665.1ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 45.8ms (= 9166.1ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.06 (= 45.8ms / 43.3ms)

OneFlow resnet50 time: 36.9ms (= 7388.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.9ms (= 7585.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.03 (= 37.9ms / 36.9ms)

OneFlow swin dataloader time: 0.251s (= 50.178s / 200, num_workers=1)
PyTorch swin dataloader time: 0.253s (= 50.560s / 200, num_workers=1)
✔️ Relative speed: 1.008 (= 0.253s / 0.251s)

OneFlow swin dataloader time: 0.064s (= 12.858s / 200, num_workers=4)
PyTorch swin dataloader time: 0.070s (= 13.980s / 200, num_workers=4)
✔️ Relative speed: 1.087 (= 0.070s / 0.064s)

OneFlow swin dataloader time: 0.037s (= 7.419s / 200, num_workers=8)
PyTorch swin dataloader time: 0.037s (= 7.357s / 200, num_workers=8)
✔️ Relative speed: 0.992 (= 0.037s / 0.037s)

✔️ OneFlow resnet50 time: 135.1ms (= 13512.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 158.2ms (= 15818.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 158.2ms / 135.1ms)

OneFlow resnet50 time: 88.0ms (= 8800.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 98.7ms (= 9866.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 98.7ms / 88.0ms)

OneFlow resnet50 time: 60.1ms (= 12014.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.7ms (= 15340.2ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 76.7ms / 60.1ms)

OneFlow resnet50 time: 50.6ms (= 10122.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 66.2ms (= 13231.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 66.2ms / 50.6ms)

OneFlow resnet50 time: 48.3ms (= 9650.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 62.7ms (= 12541.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 62.7ms / 48.3ms)

@github-actions
Copy link
Contributor

Speed stats: 
GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.4ms (= 12842.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 140.8ms (= 14078.1ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.10 (= 140.8ms / 128.4ms)

OneFlow resnet50 time: 79.8ms (= 7980.0ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 85.3ms (= 8525.7ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 85.3ms / 79.8ms)

OneFlow resnet50 time: 51.7ms (= 10349.6ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.2ms (= 11831.5ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.14 (= 59.2ms / 51.7ms)

OneFlow resnet50 time: 40.7ms (= 8130.0ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 48.3ms (= 9661.8ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.19 (= 48.3ms / 40.7ms)

OneFlow resnet50 time: 37.6ms (= 7523.1ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.6ms (= 7723.1ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.03 (= 38.6ms / 37.6ms)

OneFlow swin dataloader time: 0.249s (= 49.766s / 200, num_workers=1)
PyTorch swin dataloader time: 0.252s (= 50.368s / 200, num_workers=1)
✔️ Relative speed: 1.012 (= 0.252s / 0.249s)

OneFlow swin dataloader time: 0.067s (= 13.360s / 200, num_workers=4)
PyTorch swin dataloader time: 0.069s (= 13.830s / 200, num_workers=4)
✔️ Relative speed: 1.035 (= 0.069s / 0.067s)

OneFlow swin dataloader time: 0.037s (= 7.334s / 200, num_workers=8)
PyTorch swin dataloader time: 0.037s (= 7.343s / 200, num_workers=8)
✔️ Relative speed: 1.001 (= 0.037s / 0.037s)

✔️ OneFlow resnet50 time: 135.2ms (= 13522.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 155.2ms (= 15522.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 155.2ms / 135.2ms)

OneFlow resnet50 time: 88.7ms (= 8873.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 99.2ms (= 9917.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.12 (= 99.2ms / 88.7ms)

OneFlow resnet50 time: 60.1ms (= 12015.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 74.3ms (= 14857.5ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.24 (= 74.3ms / 60.1ms)

OneFlow resnet50 time: 53.8ms (= 10760.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.7ms (= 13946.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 69.7ms / 53.8ms)

OneFlow resnet50 time: 50.3ms (= 10062.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.1ms (= 13615.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.35 (= 68.1ms / 50.3ms)

@github-actions
Copy link
Contributor

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8043/

@github-actions
Copy link
Contributor

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

@BBuf BBuf added the automerge label Apr 19, 2022
@BBuf BBuf requested review from oneflow-ci-bot and removed request for oneflow-ci-bot April 19, 2022 03:22
@jackalcooper jackalcooper disabled auto-merge April 19, 2022 03:38
@jackalcooper jackalcooper merged commit f25e198 into master Apr 19, 2022
@jackalcooper jackalcooper deleted the alleviate_benchmark_failure branch April 19, 2022 03:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@BBuf BBuf BBuf approved these changes

@howin98 howin98 howin98 approved these changes

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

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

Successfully merging this pull request may close these issues.
3 participants
@jackalcooper @BBuf @howin98