Fix DecodeRandom reuse mem

[liujuncheng]

No description provided.

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

PyTorch resnet50 time: 152.9ms (= 7643.7ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 126.3ms (= 6314.6ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
Relative speed: 1.21 (= 152.9ms / 126.3ms)

PyTorch resnet50 time: 101.9ms (= 5095.7ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 73.8ms (= 3692.1ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
Relative speed: 1.38 (= 101.9ms / 73.8ms)

PyTorch resnet50 time: 59.4ms (= 2969.6ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 51.0ms (= 2548.1ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
Relative speed: 1.17 (= 59.4ms / 51.0ms)

PyTorch resnet50 time: 55.4ms (= 2768.8ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 54.6ms (= 2729.7ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
Relative speed: 1.01 (= 55.4ms / 54.6ms)

PyTorch resnet50 time: 44.5ms (= 2226.8ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 52.2ms (= 2611.7ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
Relative speed: 0.85 (= 44.5ms / 52.2ms)