onnx convert to engine failed

[QingL0218]

I use tensorRT8.6 trtexec.exe to convert onnx to engine ,but it failed !! The code I used is "trtexec.exe --onnx=./rtdetr_r18vd_6x_mp.onnx --workspace=4096 --shapes=image:1x3x640x640 --saveEngine=./rtdetr_r18vd_6x_mp.trt --avgRuns=100 --fp16". The error lists as follow:

I do some experiment as follow:

1. run onnx model in onnxruntime ,it can get right result;
2. use onnxruntim to get the type and shape of inputs of node "p2o.Reshape.50", the results are right

I have no idea to deal with this problem , can someone help me ??? Thx

[lyuwenyu]

具体不知道怎么回事，建议使用torch转出onnx试一下

[QingL0218]

具体不知道怎么回事，建议使用torch转出onnx试一下

我是用paddle框架下的代码训练的，用torch转换岂不是要用pytorch的训练代码？