fabio-sim
Regarding the shapes of the input and output of the model nodes
Hello, author. I'd like to ask you a question. After I converted DepthAnythingV2 to onnx using the code you provided, I wanted to view the output of each node, including the shape of the nodes and other contents. However, when I opened it with netron, I could only see the bias or weight of the node input. The shape of the data input into the node and the shape of the output of the node cannot be seen. I compared the yolo model and found that the onnx model of yolo can show the shape of the output and input of each node. What is the reason for this? May I ask, if I want to see the shape of the input and output of each node after DepthAnythingV2 is converted to onnx, how can I operate it?
我想使用 onnxsim
Thank you for your reply. Yes, your method is correct. Indeed, after using onnxsim, the output of each node can be seen. Moreover, I also tested the results of the two models processing the same image. The implementation shows that the maximum difference between them is 0. It can be proved that the model processed by onnxsim is different from the original model except for some nodes (onnxsim will delete and fold some nodes). Their results are consistent. Thank you again for your reply!