I have encountered the error “CUDNN_STATUS_INTERNAL_ERROR”, the solution I found has to reboot. Is there any way to solve it without a reboot?
The codes are similar to the following, but I couldn’t reproduce it without knowing the reason…
import torch
import torch.nn as nn
a = torch.randn([32, 1280, 7, 7])
conv1x1 = nn.Conv2d(1280, 1280, kernel_size=1, stride=1)
a = conv1x1(a)