In function importTopK: [8] Assertion failed: inputs.at(1).is_weights()

[rgc183]

from https://github.com/onnx/onnx/blob/master/docs/Operators.md#topk second input seems to be 1-D tensor with single value, however, importer for TopK expects second input to be weight and crashes.

Whats the workaround for this?

[KevenLee]

Do you solve it? By how

[simutisernestas]

@rgc183 did you solve it?

[kevinch-nv]

Hi all, TensorRT unfortunately only supports constant values for k in its TopK layer.

Are the K values in the models actually dynamic? Perhaps they can be constant-folded using the following library: https://github.com/daquexian/onnx-simplifier

[hyperfraise]

I'm also very interested in this.

[RunningLeon]

@kevinch-nv
Hi, I'm facing the same problem and K of TopK should be dynamic if we feed dynamic input into a model using TensorRT.
And in ONNXRuntime the K is dynamic.
BTW, is there any plan on this issue?
Thanks.

[wkl2013DeepVision]

Hi all, TensorRT unfortunately only supports constant values for k in its TopK layer.

Are the K values in the models actually dynamic? Perhaps they can be constant-folded using the following library: https://github.com/daquexian/onnx-simplifier

found your many answers are not the quesqtions's answers., please seriously.