extract a slice of tensor with dynamic batch size

Description

I'm trying to extract a slice from the YOLO output with a dynamic batch size (-1), but the ISliceLayer cannot accept -1 as a shape value.

auto batch_size = -1;
Dims3 shapes{batch_size, network->getOutput(0)->getDimensions().d[1], 4};
auto box_layer = network->addSlice(*network->getOutput(0), {1, 1, 1}, shapes, {0, 0, 0});
auto result = network->addShuffle(*box_layer->getOutput(0));
result->setReshapeDimensions(Dims4{batch_size, 1, 1, 4});

i got the error:
FTL: [graphShapeAnalyzer.cpp::addVolumeCheck::877] Error Code 9: Internal Error ((Unnamed Layer* 665) [Slice]_output: dimension 0 never exceeds -1)
FTL: 4: (Unnamed Layer* 665) [Slice]: mismatch in number of dimensions for start.
FTL: 4: [network.cpp::validate::3121] Error Code 4: Internal Error (Layer (Unnamed Layer* 665) [Slice] failed validation)

How can I solve this issue?

Environment

TensorRT 8.6.1.6

Cuda 12.0

Driver Version: 525.105.17

Operating System:

ubuntu 18.04