🎈permute(0, 1, 3, 4, 2),x[i]有5个维度,(2,3,4)变成(3,4,2),x(bs,3,85,20,20)to x(bs,3,20,20,85) 🎈contiguous()进行一个拷贝 然后又进入一个是否进行训练的框架中(if not self.training:) if not self.training: # inference if self.onnx_dynamic or self.grid[i].shape[2...
Hi everyone, I encounter a strange issue. I tried to convert detectron2 RoI Heads, I could convert it to ONNX successfully, but I get this error on inference: When I check the docs of onnx, the input shapes are: boxes -> [batch, N, 4], s...