TensorRT:bilinear IResizeLayer

问题1:pytorch训练过程中遇到:

UserWarning: Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details.

提示的意思是:自从pytorch0.4.0以后,upsample中当选择bilinear[双线性差值]方式时,align_corners将默认设置为False。如果有需求,可自行将其设置为True。这是正常的warning而已,不用担心。

问题2:TensorRT和Pytorch中关于align_corners设置不一致的问题

bilinear IResizeLayer setting align_corners=False makes different result to torch interpolate

为了详细的了解这个问题并解决该问题,查了一些资料,记录下来

(1)upsampling参数:resize mode + coordinate transformation mode

Here are two components, resize mode and coordinate transformation mode.
resize mode -- like nearest, bilin
上一篇:Hashmap前世今生


下一篇:asynch descriptor resize等待事件