Fix swin dataloader bug

[BBuf]

linked pr: Oneflow-Inc/models#334

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 129.0ms (= 12902.5ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 137.6ms (= 13762.6ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.07 (= 137.6ms / 129.0ms)

OneFlow resnet50 time: 79.4ms (= 7935.8ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 83.7ms (= 8374.3ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.06 (= 83.7ms / 79.4ms)

OneFlow resnet50 time: 52.0ms (= 10400.0ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 58.2ms (= 11630.3ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.12 (= 58.2ms / 52.0ms)

OneFlow resnet50 time: 43.5ms (= 8705.8ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 49.1ms (= 9816.6ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.13 (= 49.1ms / 43.5ms)

OneFlow resnet50 time: 38.7ms (= 7736.9ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.2ms (= 7445.9ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.96 (= 37.2ms / 38.7ms)

OneFlow swin dataloader time: 0.249s (= 49.752s / 200, num_workers=1)
PyTorch swin dataloader time: 0.147s (= 29.397s / 200, num_workers=1)
Relative speed: 0.591 (= 0.147s / 0.249s)

OneFlow swin dataloader time: 0.068s (= 13.546s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.234s / 200, num_workers=4)
Relative speed: 0.608 (= 0.041s / 0.068s)

OneFlow swin dataloader time: 0.036s (= 7.151s / 200, num_workers=8)
PyTorch swin dataloader time: 0.021s (= 4.258s / 200, num_workers=8)
Relative speed: 0.595 (= 0.021s / 0.036s)

✔️ OneFlow resnet50 time: 135.5ms (= 13551.4ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 163.4ms (= 16339.9ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.21 (= 163.4ms / 135.5ms)

OneFlow resnet50 time: 87.7ms (= 8769.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 99.8ms (= 9982.7ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.14 (= 99.8ms / 87.7ms)

OneFlow resnet50 time: 60.8ms (= 12168.3ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 85.5ms (= 17099.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.41 (= 85.5ms / 60.8ms)

OneFlow resnet50 time: 51.8ms (= 10359.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.4ms (= 15276.3ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.47 (= 76.4ms / 51.8ms)

OneFlow resnet50 time: 47.3ms (= 9450.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 61.8ms (= 12350.6ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.31 (= 61.8ms / 47.3ms)

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8037/

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.7ms (= 12868.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 144.6ms (= 14459.9ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.12 (= 144.6ms / 128.7ms)

OneFlow resnet50 time: 78.4ms (= 7841.7ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.3ms (= 8425.8ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.07 (= 84.3ms / 78.4ms)

OneFlow resnet50 time: 54.4ms (= 10871.7ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 60.8ms (= 12163.1ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.12 (= 60.8ms / 54.4ms)

OneFlow resnet50 time: 43.8ms (= 8751.2ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 42.4ms (= 8482.4ms / 200, input_shape=[2, 3, 224, 224])
❌ Relative speed: 0.97 (= 42.4ms / 43.8ms)

OneFlow resnet50 time: 35.3ms (= 7065.2ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.6ms (= 7722.7ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.09 (= 38.6ms / 35.3ms)

OneFlow swin dataloader time: 0.260s (= 52.011s / 200, num_workers=1)
PyTorch swin dataloader time: 0.150s (= 29.978s / 200, num_workers=1)
Relative speed: 0.576 (= 0.150s / 0.260s)

OneFlow swin dataloader time: 0.067s (= 13.393s / 200, num_workers=4)
PyTorch swin dataloader time: 0.043s (= 8.538s / 200, num_workers=4)
Relative speed: 0.637 (= 0.043s / 0.067s)

OneFlow swin dataloader time: 0.037s (= 7.496s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.535s / 200, num_workers=8)
Relative speed: 0.605 (= 0.023s / 0.037s)

✔️ OneFlow resnet50 time: 135.5ms (= 13550.3ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 162.0ms (= 16197.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 162.0ms / 135.5ms)

OneFlow resnet50 time: 87.5ms (= 8748.0ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.5ms (= 10147.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 101.5ms / 87.5ms)

OneFlow resnet50 time: 61.5ms (= 12303.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 78.6ms (= 15724.9ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.28 (= 78.6ms / 61.5ms)

OneFlow resnet50 time: 52.4ms (= 10489.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 69.5ms (= 13892.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.32 (= 69.5ms / 52.4ms)

OneFlow resnet50 time: 48.3ms (= 9665.8ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 66.7ms (= 13347.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.38 (= 66.7ms / 48.3ms)

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8037/

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8037/

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

✔️ OneFlow resnet50 time: 128.9ms (= 12888.0ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 144.8ms (= 14478.5ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.12 (= 144.8ms / 128.9ms)

OneFlow resnet50 time: 77.3ms (= 7730.8ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 83.7ms (= 8366.0ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.08 (= 83.7ms / 77.3ms)

OneFlow resnet50 time: 53.2ms (= 10648.0ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 59.0ms (= 11798.3ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.11 (= 59.0ms / 53.2ms)

OneFlow resnet50 time: 43.2ms (= 8633.1ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 40.5ms (= 8108.0ms / 200, input_shape=[2, 3, 224, 224])
❌ Relative speed: 0.94 (= 40.5ms / 43.2ms)

OneFlow resnet50 time: 38.5ms (= 7706.8ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 37.9ms (= 7573.7ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 0.98 (= 37.9ms / 38.5ms)

OneFlow swin dataloader time: 0.267s (= 53.449s / 200, num_workers=1)
PyTorch swin dataloader time: 0.151s (= 30.210s / 200, num_workers=1)
Relative speed: 0.565 (= 0.151s / 0.267s)

OneFlow swin dataloader time: 0.069s (= 13.813s / 200, num_workers=4)
PyTorch swin dataloader time: 0.042s (= 8.337s / 200, num_workers=4)
Relative speed: 0.604 (= 0.042s / 0.069s)

OneFlow swin dataloader time: 0.037s (= 7.473s / 200, num_workers=8)
PyTorch swin dataloader time: 0.023s (= 4.516s / 200, num_workers=8)
Relative speed: 0.604 (= 0.023s / 0.037s)

✔️ OneFlow resnet50 time: 135.0ms (= 13502.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 160.4ms (= 16040.1ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.19 (= 160.4ms / 135.0ms)

OneFlow resnet50 time: 89.2ms (= 8915.6ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 101.7ms (= 10172.9ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.14 (= 101.7ms / 89.2ms)

OneFlow resnet50 time: 59.5ms (= 11892.1ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 77.2ms (= 15449.7ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.30 (= 77.2ms / 59.5ms)

OneFlow resnet50 time: 51.1ms (= 10224.4ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 76.0ms (= 15197.8ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.49 (= 76.0ms / 51.1ms)

OneFlow resnet50 time: 47.7ms (= 9537.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.4ms (= 15075.3ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.58 (= 75.4ms / 47.7ms)