Remove NewKernelUtil::DnnIf

[liujuncheng]

No description provided.

[github-actions]

Speed stats:

[github-actions]

Speed stats:

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 129.5ms (= 12955.0ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 141.8ms (= 14177.5ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.09 (= 141.8ms / 129.5ms)

OneFlow resnet50 time: 79.0ms (= 7902.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 83.6ms (= 8358.9ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.06 (= 83.6ms / 79.0ms)

OneFlow resnet50 time: 52.6ms (= 10523.8ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.9ms (= 11789.8ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.12 (= 58.9ms / 52.6ms)

OneFlow resnet50 time: 40.9ms (= 8178.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 42.3ms (= 8453.3ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.03 (= 42.3ms / 40.9ms)

OneFlow resnet50 time: 36.0ms (= 7206.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.5ms (= 7500.2ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.04 (= 37.5ms / 36.0ms)

OneFlow swin dataloader time: 0.261s (= 52.276s / 200, num_workers=1)
PyTorch swin dataloader time: 0.152s (= 30.303s / 200, num_workers=1)
Relative speed: 0.580 (= 0.152s / 0.261s)

OneFlow swin dataloader time: 0.065s (= 13.050s / 200, num_workers=4)
PyTorch swin dataloader time: 0.043s (= 8.600s / 200, num_workers=4)
Relative speed: 0.659 (= 0.043s / 0.065s)

OneFlow swin dataloader time: 0.038s (= 7.587s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.419s / 200, num_workers=8)
Relative speed: 0.582 (= 0.022s / 0.038s)

❌ OneFlow resnet50 time: 144.7ms (= 14473.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 169.2ms (= 16924.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 169.2ms / 144.7ms)

OneFlow resnet50 time: 97.0ms (= 9700.2ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 111.1ms (= 11107.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 111.1ms / 97.0ms)

OneFlow resnet50 time: 77.9ms (= 15578.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 88.1ms (= 17624.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.13 (= 88.1ms / 77.9ms)

OneFlow resnet50 time: 63.3ms (= 12661.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 73.5ms (= 14702.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.16 (= 73.5ms / 63.3ms)

OneFlow resnet50 time: 57.8ms (= 11565.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.1ms (= 14422.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.25 (= 72.1ms / 57.8ms)

[github-actions]

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

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8141/