TensorRT version: 8.5.2 With JetPack 5.1.2 in AGX Orin does not support TopK from ONNX

TensorRT version: 8.5.2 With JetPack 5.1.2 in AGX Orin does not support TopK from ONNX
:
[11/16/2023-10:31:19] [E] [TRT] ModelImporter.cpp:726: While parsing node number 230 [TopK → “/extractor/TopK_output_0”]:
[11/16/2023-10:31:19] [E] [TRT] ModelImporter.cpp:727: — Begin node —
[11/16/2023-10:31:19] [E] [TRT] ModelImporter.cpp:728: input: “/extractor/GatherND_1_output_0”
input: “/extractor/Reshape_4_output_0”
output: “/extractor/TopK_output_0”
output: “/extractor/TopK_output_1”
name: “/extractor/TopK”
op_type: “TopK”
attribute {
name: “axis”
i: 0
type: INT
}
attribute {
name: “largest”
i: 1
type: INT
}
attribute {
name: “sorted”
i: 1
type: INT
}

In Windows, TensorRT 8.6.1.6 has no such problem.

Thanks!

Scott

The onnx is https://github.com/fabio-sim/LightGlue-ONNX/releases/download/v0.1.3/superpoint_2048_lightglue_end2end.onnx

Hi @chenscottusa, perhaps you can try exporting the model with a lower onnx opset, but JetPack 6 will be released later this month with TensorRT 8.6.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.