optimize ci speed in expensive eager test

[BBuf]

• AutoTest在梯度的数值无法对齐时输出梯度的具体值。
• 减少 expensive/test_convtranspose.py 运行时间。175s->12s。
• 减少 expensive/test_einsum.py 运行时间（和德澎确认过）。262s->84s。
• 减少 expensive/test_sqrt_square_sum.py 运行时间。119s->18s。

以上 expensive 里耗时巨大的Eager Test可以减少到原来总执行时间的 (175 + 262 + 119) / (12 + 84 + 18) = 556 / 114 等于4.8倍。

[github-actions]

Speed stats:

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8504/

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 129.1ms (= 12912.2ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 143.9ms (= 14388.5ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.11 (= 143.9ms / 129.1ms)

OneFlow resnet50 time: 75.4ms (= 7543.3ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.3ms (= 8425.3ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.12 (= 84.3ms / 75.4ms)

OneFlow resnet50 time: 48.8ms (= 9761.3ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.7ms (= 11940.0ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.22 (= 59.7ms / 48.8ms)

OneFlow resnet50 time: 39.9ms (= 7974.4ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 45.2ms (= 9030.7ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.13 (= 45.2ms / 39.9ms)

OneFlow resnet50 time: 35.0ms (= 6997.1ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 39.0ms (= 7794.3ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.11 (= 39.0ms / 35.0ms)

OneFlow swin dataloader time: 0.263s (= 52.554s / 200, num_workers=1)
PyTorch swin dataloader time: 0.157s (= 31.345s / 200, num_workers=1)
Relative speed: 0.596 (= 0.157s / 0.263s)

OneFlow swin dataloader time: 0.083s (= 16.618s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.240s / 200, num_workers=4)
Relative speed: 0.496 (= 0.041s / 0.083s)

OneFlow swin dataloader time: 0.045s (= 8.929s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.672s / 200, num_workers=8)
Relative speed: 0.523 (= 0.023s / 0.045s)

❌ OneFlow resnet50 time: 144.3ms (= 14427.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 169.2ms (= 16918.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 169.2ms / 144.3ms)

OneFlow resnet50 time: 92.5ms (= 9248.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 116.5ms (= 11647.8ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.26 (= 116.5ms / 92.5ms)

OneFlow resnet50 time: 68.6ms (= 13720.0ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 89.0ms (= 17791.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 89.0ms / 68.6ms)

OneFlow resnet50 time: 57.2ms (= 11433.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 81.2ms (= 16231.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.42 (= 81.2ms / 57.2ms)

OneFlow resnet50 time: 52.9ms (= 10575.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.3ms (= 13865.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 69.3ms / 52.9ms)

[github-actions]

Speed stats:

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8504/

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

✔️ OneFlow resnet50 time: 129.0ms (= 12903.4ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 142.9ms (= 14285.3ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.11 (= 142.9ms / 129.0ms)

OneFlow resnet50 time: 75.5ms (= 7552.5ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.7ms (= 8471.1ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.12 (= 84.7ms / 75.5ms)

OneFlow resnet50 time: 49.4ms (= 9889.9ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 57.9ms (= 11577.8ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.17 (= 57.9ms / 49.4ms)

OneFlow resnet50 time: 39.9ms (= 7973.7ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 44.6ms (= 8930.0ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.12 (= 44.6ms / 39.9ms)

OneFlow resnet50 time: 36.1ms (= 7215.6ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.8ms (= 7567.8ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.05 (= 37.8ms / 36.1ms)

OneFlow swin dataloader time: 0.286s (= 57.147s / 200, num_workers=1)
PyTorch swin dataloader time: 0.155s (= 30.915s / 200, num_workers=1)
Relative speed: 0.541 (= 0.155s / 0.286s)

OneFlow swin dataloader time: 0.082s (= 16.410s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.137s / 200, num_workers=4)
Relative speed: 0.496 (= 0.041s / 0.082s)

OneFlow swin dataloader time: 0.047s (= 9.410s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.560s / 200, num_workers=8)
Relative speed: 0.485 (= 0.023s / 0.047s)

❌ OneFlow resnet50 time: 145.3ms (= 14532.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 170.7ms (= 17066.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 170.7ms / 145.3ms)

OneFlow resnet50 time: 93.8ms (= 9376.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 114.0ms (= 11399.1ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 114.0ms / 93.8ms)

OneFlow resnet50 time: 70.4ms (= 14070.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 95.8ms (= 19164.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.36 (= 95.8ms / 70.4ms)

OneFlow resnet50 time: 57.0ms (= 11403.0ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.9ms (= 15172.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.33 (= 75.9ms / 57.0ms)

OneFlow resnet50 time: 51.0ms (= 10208.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 79.1ms (= 15827.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.55 (= 79.1ms / 51.0ms)