Add cast primitive

[liujuncheng]

No description provided.

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 128.9ms (= 6447.2ms / 50, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 139.1ms (= 6954.7ms / 50, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.08 (= 139.1ms / 128.9ms)

OneFlow resnet50 time: 75.3ms (= 3765.3ms / 50, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 83.3ms (= 4166.0ms / 50, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.11 (= 83.3ms / 75.3ms)

OneFlow resnet50 time: 50.0ms (= 2499.1ms / 50, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.7ms (= 2987.1ms / 50, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.20 (= 59.7ms / 50.0ms)

OneFlow resnet50 time: 43.4ms (= 2169.7ms / 50, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 50.9ms (= 2543.6ms / 50, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.17 (= 50.9ms / 43.4ms)

OneFlow resnet50 time: 49.0ms (= 2448.1ms / 50, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 45.5ms (= 2276.9ms / 50, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.93 (= 45.5ms / 49.0ms)

OneFlow resnet50 time: 161.7ms (= 8086.4ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 163.8ms (= 8190.2ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.01 (= 163.8ms / 161.7ms)

OneFlow resnet50 time: 100.9ms (= 5046.7ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 108.4ms (= 5422.4ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.07 (= 108.4ms / 100.9ms)

OneFlow resnet50 time: 78.8ms (= 3938.5ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 81.4ms (= 4069.4ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.03 (= 81.4ms / 78.8ms)

OneFlow resnet50 time: 70.4ms (= 3521.6ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 66.3ms (= 3316.0ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.94 (= 66.3ms / 70.4ms)

OneFlow resnet50 time: 76.1ms (= 3803.1ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 60.4ms (= 3019.1ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.79 (= 60.4ms / 76.1ms)