Table Of Contents
The batchTilePlugin tiles a tensor N times along its first dimension (batch dimension) where N is the batch size. The result tensor will have shape N on its first dimension and each tensor[i: i+1,...] is a copy of input tensor (for integer i < N).
This plugin takes 2 input tensors, the first input tensor should have first dim N and the second tensor should have first dim 1. The output tensor will be the tile results of the second tensor.
Both input tensors to this plugin must have 4 dimensions (NCHW)
For example, you have input tensor A of shape [2, 2, 2, 2]:
[[[[ 0 1]
[ 2 3]]
[[ 4 5]
[ 6 7]]]
[[[ 8 9]
[10 11]]
[[12 13]
[14 15]]]]
and input tensor B of shape [1, 3, 2, 2]:
[[[[16 17]
[18 19]]
[[20 21]
[22 23]]
[[24 25]
[26 27]]]]
After batchTilePlugin for the two inputs, the output tensor of shape [2, 3, 2, 2] is:
[[[[16 17]
[18 19]]
[[20 21]
[22 23]]
[[24 25]
[26 27]]]
[[[16 17]
[18 19]]
[[20 21]
[22 23]]
[[24 25]
[26 27]]]]
This plugin has the plugin creator class BatchTilePluginCreator and the plugin class BatchTilePlugin.
No parameter is required for this plugin.
For terms and conditions for use, reproduction, and distribution, see the TensorRT Software License Agreement documentation.
Jul. 2019
This is the first release of this README.md file.
There are no known issues in this plugin.