starimeL/PytorchConverter

pytorch pooling mode?

Closed this issue · 1 comments

Do we must use 'ceil' mode pooling, dose ncnn support 'floor' mode?

Mind the difference on ceil_mode of pooling layer among Pytorch and Caffe, ncnn

  • You can convert Pytorch models with all pooling layer's ceil_mode=True.
  • Or compile a custom version of Caffe/ncnn with floor() replaced by ceil() in pooling layer inference.