search

NVIDIA / cudnn-frontend

cudnn_frontend provides a c++ wrapper for the cudnn backend API and samples on how to use it
MIT License
424 stars 86 forks source link

question about memory layout of the convolution #83

Closed jinz2014 closed 2 months ago

jinz2014 commented 3 months ago

How could I disable channel-last memory layout in the graph.conv_fprop(...) function ? So the layout is NCHW instead. Thanks.

Anerudhan commented 3 months ago

Hi @jinz2014 ,

Thanks for your question

You can manually set the output stride order. See here. This sample has channel first throughout.