When I use TensorRT to convert model with over 100 custom layers, the registerCreator() API gives error: Cannot register creator - Max plugin limit reached. Current limit is 100.
Is there any way to explicitly set the maximum plugin limit?
I have read the " Estimating Depth with ONNX" blog before, and the problem is this method must re-compile the TRT-related libraries, which is not what we expected.
For this “Plugin limit” issue, I think it’s not related with how to implement custom plugins, but something about how to set the parameter of “maximum plugin limit”. What I’m concerned about is whether I could set this parameter so that I can use more than 100 custom plugins in one model, or this parameter is already fixed and cannot be changed(which means that I could only use <=100 custom plugins in one model).