Update dependency constraints on cudnn

[jackalcooper]

No description provided.

[github-actions]

View latest API docs preview at: https://oneflow-staging.oss-cn-beijing.aliyuncs.com/docs/Oneflow-Inc/oneflow/pr/10515/

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce RTX 3080 Ti 

❌ OneFlow resnet50 time: 43.9ms (= 4385.3ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 57.6ms (= 5762.7ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.31 (= 57.6ms / 43.9ms)

OneFlow resnet50 time: 26.5ms (= 2647.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 37.8ms (= 3776.0ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.43 (= 37.8ms / 26.5ms)

OneFlow resnet50 time: 18.5ms (= 3699.5ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 35.1ms (= 7029.0ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.90 (= 35.1ms / 18.5ms)

OneFlow resnet50 time: 17.3ms (= 3451.0ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 31.5ms (= 6292.7ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.82 (= 31.5ms / 17.3ms)

OneFlow resnet50 time: 17.2ms (= 3436.8ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 31.0ms (= 6190.6ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.80 (= 31.0ms / 17.2ms)

OneFlow swin dataloader time: 0.201s (= 40.170s / 200, num_workers=1)
PyTorch swin dataloader time: 0.128s (= 25.600s / 200, num_workers=1)
Relative speed: 0.637 (= 0.128s / 0.201s)

OneFlow swin dataloader time: 0.055s (= 10.995s / 200, num_workers=4)
PyTorch swin dataloader time: 0.033s (= 6.586s / 200, num_workers=4)
Relative speed: 0.599 (= 0.033s / 0.055s)

OneFlow swin dataloader time: 0.031s (= 6.162s / 200, num_workers=8)
PyTorch swin dataloader time: 0.017s (= 3.331s / 200, num_workers=8)
Relative speed: 0.541 (= 0.017s / 0.031s)

❌ OneFlow resnet50 time: 49.3ms (= 4929.2ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 64.4ms (= 6444.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 64.4ms / 49.3ms)

OneFlow resnet50 time: 36.1ms (= 3610.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 45.9ms (= 4588.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.27 (= 45.9ms / 36.1ms)

OneFlow resnet50 time: 27.5ms (= 5502.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 38.5ms (= 7702.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.40 (= 38.5ms / 27.5ms)

OneFlow resnet50 time: 25.1ms (= 5024.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 40.7ms (= 8139.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.62 (= 40.7ms / 25.1ms)

OneFlow resnet50 time: 24.8ms (= 4962.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 36.1ms (= 7227.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.46 (= 36.1ms / 24.8ms)