add protobuf < 4.0

[jackalcooper]

No description provided.

[shangguanshiyuan]

我试了protobuf-4.0.0rc1和protobuf-4.0.0rc2不报错，目前4.0就这三个版本，只有4.21.0rc1有错。是只排除4.21.0rc1还是选择暂时把4.0都排除？

[jackalcooper]

我试了protobuf-4.0.0rc1和protobuf-4.0.0rc2不报错，目前4.0就这三个版本，只有4.21.0rc1有错。是只排除4.21.0rc1还是选择暂时把4.0都排除？

暂时把4.0都排除吧，除了 4.21.0rc1 发现的问题，跨大版本大概率还有潜在的问题。我们的c++是3的。

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8198/

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 129.2ms (= 12918.3ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 150.0ms (= 14996.7ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.16 (= 150.0ms / 129.2ms)

OneFlow resnet50 time: 78.3ms (= 7828.5ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 83.9ms (= 8393.2ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 83.9ms / 78.3ms)

OneFlow resnet50 time: 52.4ms (= 10489.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 63.2ms (= 12649.1ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.21 (= 63.2ms / 52.4ms)

OneFlow resnet50 time: 41.3ms (= 8264.2ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 41.2ms (= 8242.4ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.00 (= 41.2ms / 41.3ms)

OneFlow resnet50 time: 36.0ms (= 7206.7ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 39.6ms (= 7926.4ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.10 (= 39.6ms / 36.0ms)

OneFlow swin dataloader time: 0.250s (= 50.018s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.258s / 200, num_workers=1)
Relative speed: 0.605 (= 0.151s / 0.250s)

OneFlow swin dataloader time: 0.064s (= 12.898s / 200, num_workers=4)
PyTorch swin dataloader time: 0.043s (= 8.627s / 200, num_workers=4)
Relative speed: 0.669 (= 0.043s / 0.064s)

OneFlow swin dataloader time: 0.036s (= 7.238s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.389s / 200, num_workers=8)
Relative speed: 0.606 (= 0.022s / 0.036s)

❌ OneFlow resnet50 time: 145.0ms (= 14499.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 169.6ms (= 16955.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.17 (= 169.6ms / 145.0ms)

OneFlow resnet50 time: 97.4ms (= 9739.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 110.4ms (= 11035.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.13 (= 110.4ms / 97.4ms)

OneFlow resnet50 time: 73.6ms (= 14712.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 90.2ms (= 18037.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.23 (= 90.2ms / 73.6ms)

OneFlow resnet50 time: 63.2ms (= 12636.1ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 73.2ms (= 14648.7ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
❌ Relative speed: 1.16 (= 73.2ms / 63.2ms)

OneFlow resnet50 time: 57.1ms (= 11421.4ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.6ms (= 13922.5ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 69.6ms / 57.1ms)