Optim upsample backward

[BBuf]

UpsampleNearesrt2D在采样倍率为2的时候不使用NdIndexOffsetHelper，减少计算量。

Pack dy减少访存，避免使用atomicAdd。

[github-actions]

Code got formatted by CI. Please request CI again if you still want to have this PR merged. If the PR is from a forked repo, please download the patch files from the GitHub Actions web page and apply them locally.

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 









❌ OneFlow resnet50 time: 140.1ms (= 14009.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 165.2ms (= 16518.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.18 (= 165.2ms / 140.1ms)

OneFlow resnet50 time: 85.2ms (= 8522.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 102.8ms (= 10283.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.21 (= 102.8ms / 85.2ms)

OneFlow resnet50 time: 57.8ms (= 11560.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 81.5ms (= 16309.6ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.41 (= 81.5ms / 57.8ms)

OneFlow resnet50 time: 44.7ms (= 8944.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.3ms (= 14450.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.62 (= 72.3ms / 44.7ms)

OneFlow resnet50 time: 40.0ms (= 8006.0ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 67.9ms (= 13571.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.70 (= 67.9ms / 40.0ms)

[github-actions]

CI failed when running job: cuda-misc. PR label automerge has been removed

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 









❌ OneFlow resnet50 time: 139.9ms (= 13987.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 160.6ms (= 16063.0ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.15 (= 160.6ms / 139.9ms)

OneFlow resnet50 time: 84.9ms (= 8492.3ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.9ms (= 10190.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 101.9ms / 84.9ms)

OneFlow resnet50 time: 57.6ms (= 11520.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 80.1ms (= 16012.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.39 (= 80.1ms / 57.6ms)

OneFlow resnet50 time: 45.8ms (= 9155.9ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 72.1ms (= 14425.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.58 (= 72.1ms / 45.8ms)

OneFlow resnet50 time: 41.2ms (= 8248.1ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.3ms (= 13851.2ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.68 (= 69.3ms / 41.2ms)

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/9424/