Is my understanding correct that we are doing only consistency checks between cpu , gpu and cudnn runs for the bilinear sampler and removing the correctness tests that we had earlier (bilinear sampler forward and backward computation using numpy ops) ? This should be fine assuming the correctness of cudnn op.
[ Full content available at: https://github.com/apache/incubator-mxnet/pull/12427 ] This message was relayed via gitbox.apache.org for [email protected]
