speshowBUAA/mmdet3d_onnx_tools

when onnx2tensorRT, for tok op has some ERROR

melon0504 opened this issue · 5 comments

ERROR: builtin_op_importers.cpp:3345 In function importTopK:
[8] Assertion failed: inputs.at(1).is_weights()
ERROR: failed to parse onnx file

I used onnx-tensorrt to fix the problem,but now here has new error.

[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from PointPillars
[ RUN ] PointPillars.build_model
| ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄>
| ../model/pts_pfe.trt >
|_____________________>
(_/) ||
(•ㅅ•) ||
/ づ
Binding 0 (input.1): Input.
Binding 1 (38): Output.
| ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄>
| ../model/pts_backbone_with_anchors.trt >
|_____________________>
(_
/) ||
(•ㅅ•) ||
/ づ
Binding 0 (0): Input.
Binding 1 (225): Output.
Binding 2 (227): Output.
Binding 3 (273): Output.
temp_points.size() :5040001
ERROR: ../rtSafe/cuda/genericReformat.cu (1294) - Cuda Error in executeMemcpy: 1 (invalid argument)
ERROR: FAILED_EXECUTION: std::exception
ERROR: ../rtSafe/cuda/genericReformat.cu (1294) - Cuda Error in executeMemcpy: 1 (invalid argument)
ERROR: FAILED_EXECUTION: std::exception
/home/ruiming/ROS/PointPillars_mmdet_secfpn/test/pointpillars_test.cc:151: Failure
Expected equality of these values:
num_objects
Which is: 0
228
138.28 ms
ERROR: ../rtSafe/cuda/cudaScaleRunner.cpp (119) - Cuda Error in execute: 1 (invalid argument)
ERROR: FAILED_EXECUTION: std::exception
ERROR: ../rtSafe/cuda/genericReformat.cu (1294) - Cuda Error in executeMemcpy: 1 (invalid argument)
ERROR: FAILED_EXECUTION: std::exception
/home/ruiming/ROS/PointPillars_mmdet_secfpn/test/pointpillars_test.cc:151: Failure
Expected equality of these values:
num_objects
Which is: 0
228
19.93 ms
ERROR: ../rtSafe/cuda/cudaScaleRunner.cpp (119) - Cuda Error in execute: 1 (invalid argument)
ERROR: FAILED_EXECUTION: std::exception
ERROR: ../rtSafe/cuda/genericReformat.cu (1294) - Cuda Error in executeMemcpy: 1 (invalid argument)

perharps there are conficts between different tensorrt version. Try onnx model file as well.

可以加下QQ(758341407),或者微信咨询一下吗

可以