some op test

[xiezipeng-ML]

• oneflow.cat
• oneflow.nn.Dropout
• oneflow.new_ones
• oneflow.Tensor.diag
• oneflow.diag

[BBuf]

测一下p!=0&&p!=1的随机值

[xiezipeng-ML]

这个dropout随机drop其tensor，所以没办法确保pytorch和oneflow的tensor被相同的drop，只能测试p=0和p=1的情况

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

OneFlow resnet50 time: 128.6ms (= 6431.5ms / 50, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 141.1ms (= 7056.6ms / 50, input_shape=[16, 3, 224, 224])
Relative speed: 1.10 (= 141.1ms / 128.6ms)

OneFlow resnet50 time: 74.8ms (= 3737.9ms / 50, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.9ms (= 4247.4ms / 50, input_shape=[8, 3, 224, 224])
Relative speed: 1.14 (= 84.9ms / 74.8ms)

OneFlow resnet50 time: 48.8ms (= 2441.2ms / 50, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 56.9ms (= 2844.7ms / 50, input_shape=[4, 3, 224, 224])
Relative speed: 1.17 (= 56.9ms / 48.8ms)

OneFlow resnet50 time: 41.9ms (= 2097.3ms / 50, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 46.8ms (= 2341.4ms / 50, input_shape=[2, 3, 224, 224])
Relative speed: 1.12 (= 46.8ms / 41.9ms)

OneFlow resnet50 time: 41.3ms (= 2064.7ms / 50, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 44.3ms (= 2216.8ms / 50, input_shape=[1, 3, 224, 224])
Relative speed: 1.07 (= 44.3ms / 41.3ms)

OneFlow resnet50 time: 154.0ms (= 7698.9ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 166.5ms (= 8326.6ms / 50, input_shape=[16, 3, 224, 224], ddp, world size=2)
Relative speed: 1.08 (= 166.5ms / 154.0ms)

OneFlow resnet50 time: 96.9ms (= 4843.3ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 107.5ms (= 5372.8ms / 50, input_shape=[8, 3, 224, 224], ddp, world size=2)
Relative speed: 1.11 (= 107.5ms / 96.9ms)

OneFlow resnet50 time: 67.7ms (= 3386.1ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 83.1ms (= 4153.5ms / 50, input_shape=[4, 3, 224, 224], ddp, world size=2)
Relative speed: 1.23 (= 83.1ms / 67.7ms)

OneFlow resnet50 time: 61.7ms (= 3086.3ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.7ms (= 3483.3ms / 50, input_shape=[2, 3, 224, 224], ddp, world size=2)
Relative speed: 1.13 (= 69.7ms / 61.7ms)

OneFlow resnet50 time: 60.1ms (= 3006.9ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 66.1ms (= 3305.1ms / 50, input_shape=[1, 3, 224, 224], ddp, world size=2)
Relative speed: 1.10 (= 66.1ms / 60.1ms)