pp_liteseg模型池化层疑问

[enemy1205]

问题确认 Search before asking

• [X] 我已经搜索过问题，但是没有找到解答。I have searched the question and found no related answer.

请提出你的问题 Please ask your question

使用https://github.com/PaddlePaddle/PaddleSlim/blob/9f3e9b2f0f9948b780900d1299f2c3fe47322deb/example/auto_compression/semantic_segmentation/README.md 提供的pp_liteseg模型 使用paddle2onnx --model_dir RES-paddle2-PPLIteSegSTDC1 --model_filename model.pdmodel --params_filename model.pdiparams --opset_version 13 --save_file output.onnx命令转为onnx模型后发现

Node name: p2o.AveragePool.0
Kernel shape: [3, 3]
Strides: [2, 2]
Pads: [1, 1, 1, 1]
Node name: p2o.AveragePool.1
Kernel shape: [3, 3]
Strides: [2, 2]
Pads: [1, 1, 1, 1]
Node name: p2o.AveragePool.2
Kernel shape: [3, 3]
Strides: [2, 2]
Pads: [1, 1, 1, 1]
Node name: p2o.AveragePool.3
Kernel shape: [-1, -1]
Strides: [0, 0]
Node name: p2o.AveragePool.4
Kernel shape: [-1, -1]
Strides: [0, 0]

其中p2o.AveragePool.3和p2o.AveragePool.4层的kernel shape为-1 为什么AveragePool 层的 kernel_shape 属性为负值呢