Optimize transpose identity

[liujuncheng]

优化诸如形状 [a, 1, b] , perm [0, 2, 1] 的情形

[github-actions[bot]]

Speed stats:

GPU Name: GeForce GTX 1080 









❌ OneFlow resnet50 time: 140.2ms (= 14017.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 161.6ms (= 16156.8ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 161.6ms / 140.2ms)

OneFlow resnet50 time: 85.2ms (= 8519.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 104.4ms (= 10444.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.23 (= 104.4ms / 85.2ms)

OneFlow resnet50 time: 57.6ms (= 11514.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 88.4ms (= 17678.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.54 (= 88.4ms / 57.6ms)

OneFlow resnet50 time: 44.1ms (= 8828.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 71.8ms (= 14356.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.63 (= 71.8ms / 44.1ms)

OneFlow resnet50 time: 41.2ms (= 8235.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.6ms (= 13512.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.64 (= 67.6ms / 41.2ms)

[github-actions[bot]]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9416/

[github-actions[bot]]

Speed stats:

[github-actions[bot]]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 









❌ OneFlow resnet50 time: 147.6ms (= 14762.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 170.0ms (= 16995.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 170.0ms / 147.6ms)

OneFlow resnet50 time: 94.9ms (= 9486.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 113.5ms (= 11352.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 113.5ms / 94.9ms)

OneFlow resnet50 time: 68.4ms (= 13679.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 87.8ms (= 17555.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 87.8ms / 68.4ms)

OneFlow resnet50 time: 59.8ms (= 11968.6ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.0ms (= 14999.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.25 (= 75.0ms / 59.8ms)

OneFlow resnet50 time: 54.2ms (= 10834.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.4ms (= 14475.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.34 (= 72.4ms / 54.2ms)

[github-actions[bot]]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9416/