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.

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

Use unary primitive in kernel #8270

Merged
merged 16 commits into from
May 24, 2022
Merged

Use unary primitive in kernel #8270

merged 16 commits into from
May 24, 2022

Conversation

EsdeathYZH
Copy link
Contributor

No description provided.

@EsdeathYZH EsdeathYZH requested a review from liujuncheng as a code owner May 20, 2022 06:43
@EsdeathYZH EsdeathYZH requested a review from MARD1NO May 20, 2022 06:44
@EsdeathYZH EsdeathYZH requested a review from guo-ran as a code owner May 23, 2022 13:47
@github-actions
Copy link
Contributor

Speed stats:

@github-actions
Copy link
Contributor

Speed stats:
GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 130.4ms (= 13036.3ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 144.2ms (= 14417.4ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.11 (= 144.2ms / 130.4ms)

OneFlow resnet50 time: 76.8ms (= 7683.5ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 87.4ms (= 8736.9ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.14 (= 87.4ms / 76.8ms)

OneFlow resnet50 time: 54.8ms (= 10962.4ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 53.9ms (= 10774.5ms / 200, input_shape=[4, 3, 224, 224])
❌ Relative speed: 0.98 (= 53.9ms / 54.8ms)

OneFlow resnet50 time: 41.6ms (= 8317.4ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 41.2ms (= 8244.0ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 0.99 (= 41.2ms / 41.6ms)

OneFlow resnet50 time: 35.3ms (= 7059.2ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.6ms (= 7522.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.07 (= 37.6ms / 35.3ms)

OneFlow swin dataloader time: 0.247s (= 49.458s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.135s / 200, num_workers=1)
Relative speed: 0.609 (= 0.151s / 0.247s)

OneFlow swin dataloader time: 0.066s (= 13.132s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.120s / 200, num_workers=4)
Relative speed: 0.618 (= 0.041s / 0.066s)

OneFlow swin dataloader time: 0.059s (= 11.845s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.353s / 200, num_workers=8)
Relative speed: 0.367 (= 0.022s / 0.059s)

❌ OneFlow resnet50 time: 146.4ms (= 14640.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 171.1ms (= 17113.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 171.1ms / 146.4ms)

OneFlow resnet50 time: 98.8ms (= 9884.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 111.8ms (= 11180.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.13 (= 111.8ms / 98.8ms)

OneFlow resnet50 time: 74.3ms (= 14856.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 89.8ms (= 17969.1ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.21 (= 89.8ms / 74.3ms)

OneFlow resnet50 time: 60.6ms (= 12127.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 74.2ms (= 14831.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 74.2ms / 60.6ms)

OneFlow resnet50 time: 54.1ms (= 10823.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.4ms (= 13670.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.26 (= 68.4ms / 54.1ms)

@mergify mergify bot merged commit 6171e55 into master May 24, 2022
@mergify mergify bot deleted the use_unary_primitive branch May 24, 2022 11:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants