Refactor to local

[clackhan]

• 重构CastToConsistent后向，在后向过程中自动添加boxing
• 解决CastFromConsistent后向过程没有dtype attr的bug
• 删除 oneflow/core/autograd/gradient_funcs/hierarchical_parallel_cast.cpp，由于ConsistentToConsistentOpExpr有注册专门的后向function，此文件不在需要，故删除

[lixinqi]

删掉这里的理由是什么呢？

[clackhan]

删掉这里的理由是什么呢？

这里本来就是后向过程，不需要在此处配置后向的后向，可参考：https://github.com/Oneflow-Inc/oneflow/blob/master/oneflow/user/ops/hierarchical_parallel_cast_op.cpp#L92

[github-actions]

CI failed, removing label automerge

[github-actions]

CI failed, removing label automerge

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 127.5ms (= 6375.8ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 140.0ms (= 6999.6ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.10 (= 140.0ms / 127.5ms)

OneFlow resnet50 time: 74.3ms (= 3716.9ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 83.5ms (= 4176.6ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.12 (= 83.5ms / 74.3ms)

OneFlow resnet50 time: 48.4ms (= 2420.0ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 58.4ms (= 2920.5ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.21 (= 58.4ms / 48.4ms)

OneFlow resnet50 time: 40.5ms (= 2024.6ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 47.2ms (= 2359.5ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.17 (= 47.2ms / 40.5ms)

OneFlow resnet50 time: 38.2ms (= 1909.2ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 44.2ms (= 2211.9ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.16 (= 44.2ms / 38.2ms)

OneFlow resnet50 time: 145.8ms (= 7288.4ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 152.8ms (= 7640.2ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.05 (= 152.8ms / 145.8ms)

OneFlow resnet50 time: 92.9ms (= 4644.0ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 91.3ms (= 4567.0ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 0.98 (= 91.3ms / 92.9ms)

OneFlow resnet50 time: 65.8ms (= 3291.7ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 67.3ms (= 3366.7ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 1.02 (= 67.3ms / 65.8ms)

OneFlow resnet50 time: 61.5ms (= 3077.3ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 53.7ms (= 2683.9ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 0.87 (= 53.7ms / 61.5ms)

OneFlow resnet50 time: 55.2ms (= 2757.7ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow GPU used (rank 0): 0 MiB
PyTorch resnet50 time: 49.1ms (= 2456.3ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
PyTorch GPU used (rank 0, estimated): 0 MiB
Relative speed: 0.89 (= 49.1ms / 55.2ms)