Fix blob desc segfault

[jackalcooper]

No description provided.

[github-actions]

View latest API docs preview at: https://staging.oneflow.info/docs/Oneflow-Inc/oneflow/pr/8252/

[github-actions]

Speed stats:

GPU Name: NVIDIA GeForce GTX 1080 

❌ OneFlow resnet50 time: 129.7ms (= 12970.9ms / 100, input_shape=[16, 3, 224, 224])
PyTorch resnet50 time: 143.0ms (= 14300.3ms / 100, input_shape=[16, 3, 224, 224])
✔️ Relative speed: 1.10 (= 143.0ms / 129.7ms)

OneFlow resnet50 time: 77.5ms (= 7751.1ms / 100, input_shape=[8, 3, 224, 224])
PyTorch resnet50 time: 84.8ms (= 8481.3ms / 100, input_shape=[8, 3, 224, 224])
✔️ Relative speed: 1.09 (= 84.8ms / 77.5ms)

OneFlow resnet50 time: 54.1ms (= 10814.4ms / 200, input_shape=[4, 3, 224, 224])
PyTorch resnet50 time: 62.5ms (= 12509.2ms / 200, input_shape=[4, 3, 224, 224])
✔️ Relative speed: 1.16 (= 62.5ms / 54.1ms)

OneFlow resnet50 time: 40.4ms (= 8087.9ms / 200, input_shape=[2, 3, 224, 224])
PyTorch resnet50 time: 49.0ms (= 9795.9ms / 200, input_shape=[2, 3, 224, 224])
✔️ Relative speed: 1.21 (= 49.0ms / 40.4ms)

OneFlow resnet50 time: 35.3ms (= 7063.3ms / 200, input_shape=[1, 3, 224, 224])
PyTorch resnet50 time: 38.9ms (= 7774.0ms / 200, input_shape=[1, 3, 224, 224])
✔️ Relative speed: 1.10 (= 38.9ms / 35.3ms)

OneFlow swin dataloader time: 0.247s (= 49.362s / 200, num_workers=1)
PyTorch swin dataloader time: 0.152s (= 30.367s / 200, num_workers=1)
Relative speed: 0.615 (= 0.152s / 0.247s)

OneFlow swin dataloader time: 0.064s (= 12.849s / 200, num_workers=4)
PyTorch swin dataloader time: 0.041s (= 8.200s / 200, num_workers=4)
Relative speed: 0.638 (= 0.041s / 0.064s)

OneFlow swin dataloader time: 0.055s (= 11.042s / 200, num_workers=8)
PyTorch swin dataloader time: 0.022s (= 4.344s / 200, num_workers=8)
Relative speed: 0.393 (= 0.022s / 0.055s)

❌ OneFlow resnet50 time: 145.4ms (= 14540.5ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 168.6ms (= 16855.7ms / 100, input_shape=[16, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 168.6ms / 145.4ms)

OneFlow resnet50 time: 97.5ms (= 9754.4ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 112.8ms (= 11281.5ms / 100, input_shape=[8, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.16 (= 112.8ms / 97.5ms)

OneFlow resnet50 time: 72.6ms (= 14512.8ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 88.8ms (= 17752.4ms / 200, input_shape=[4, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.22 (= 88.8ms / 72.6ms)

OneFlow resnet50 time: 62.5ms (= 12493.5ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 75.2ms (= 15041.2ms / 200, input_shape=[2, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.20 (= 75.2ms / 62.5ms)

OneFlow resnet50 time: 55.4ms (= 11086.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
PyTorch resnet50 time: 68.6ms (= 13713.9ms / 200, input_shape=[1, 3, 224, 224], ddp, world size=2)
✔️ Relative speed: 1.24 (= 68.6ms / 55.4ms)