TensorRT 6.0 Global PaddingMode

hi,
I use setPoolingOutputDimensionsFormula to compute Pooling output shape in TensorRT 5.1.
However, I cann’t find similar function for this purpose in TensorRT 6.0.1 .
There ONLY have PaddingMode, but I need a method for GLOBAL setting.
Anyone have any idea about this?

Hi,
setPoolingOutputDimensionsFormula is TRT_DEPRECATED in TRT 6.0 onwards.
https://docs.nvidia.com/deeplearning/sdk/tensorrt-archived/tensorrt-601/tensorrt-api/c_api/classnvinfer1_1_1_i_network_definition.html#a402ad2fbbf572f592077c6c0ea12a51f

Currently we are providing six padding modes, which can cover most formula for DL Frameworks.
All formula is listed in below link:
https://docs.nvidia.com/deeplearning/sdk/tensorrt-api/c_api/namespacenvinfer1.html#af0cf8e1034112a1472f3a4bc00f5de62

Thanks

Hi, SunilJB,

Thanks for your reply.

I need to set PaddingMode for all layers.
I found API setPaddingMode() only for a layer, but I need a function for all layers.

Actually, I use caffeparser to convert caffe model to TRT.
And there is only one pooling mode(ceil or floor) in caffe.
Default pooling mode of TRT is ceil.
If I need change mode for all layers, how should I do?

Thanks.

Hi,

From TRT 6.0 we don’t have a global API to set global padding mode but you can use setPaddingMode() API to set it layer by layer.

Thanks