graph support eager lrs

[strint]

增加一个graph支持的lr的流程

1. c++ lr kernel
2. lr proto
3. python eager lr add graph conf
4. test in test_graph_lrs.py

待支持的

• multistep lr
• step lr
• cosine annealing lr

除了lambda lr，当前eager的lr，graph都已经支持

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 128.0ms (= 6399.0ms / 50, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 138.3ms (= 6915.6ms / 50, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.08 (= 138.3ms / 128.0ms)

OneFlow resnet50 time: 74.1ms (= 3703.8ms / 50, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 85.0ms (= 4251.7ms / 50, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.15 (= 85.0ms / 74.1ms)

OneFlow resnet50 time: 49.4ms (= 2471.7ms / 50, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.1ms (= 2953.9ms / 50, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.20 (= 59.1ms / 49.4ms)

OneFlow resnet50 time: 43.7ms (= 2187.4ms / 50, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 49.4ms (= 2467.7ms / 50, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.13 (= 49.4ms / 43.7ms)

OneFlow resnet50 time: 44.3ms (= 2212.8ms / 50, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 43.8ms (= 2189.8ms / 50, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.99 (= 43.8ms / 44.3ms)

OneFlow resnet50 time: 160.5ms (= 8023.1ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 158.3ms (= 7915.7ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.99 (= 158.3ms / 160.5ms)

OneFlow resnet50 time: 103.0ms (= 5149.3ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.3ms (= 5117.1ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.99 (= 102.3ms / 103.0ms)

OneFlow resnet50 time: 80.3ms (= 4016.2ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 74.6ms (= 3729.2ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.93 (= 74.6ms / 80.3ms)

OneFlow resnet50 time: 77.6ms (= 3882.3ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 63.8ms (= 3187.8ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.82 (= 63.8ms / 77.6ms)

OneFlow resnet50 time: 69.4ms (= 3469.8ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 60.3ms (= 3013.5ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 0.87 (= 60.3ms / 69.4ms)