Caffe Dropout layers

Hello,

In Caffe, Dropout layers are automatically removed in the test phase defined by

caffe::Net<float>(modelFile, caffe::TEST)

since they are only usefull during the training phase.

When optimizing a caffe model to TRT engine with the caffe parser, are the Dropout layers automatically removed ?

Hi,

TRT ignores layers marked as TEST when parsing the caffe protobuf.

Thanks