deep learning - Where is the implementation of Convolution of tensorflow -
how tensorflow implements convolution operations? know it's based on eigen library , trace launchgeneric.launch
in /tensorflow/core/kernels/conv_ops.cc, goes? guess in /third_party_eigen3/unsupported/eigen/cxx11/src/neuralnetworks/spatialconvolutions.h, checked build file under eigen3 , seems file not included compile library.
the gpu implementation cudnn. think you're right on cpu implementation, , it's included in build eigen's build file via "eigen/**"
(recursive wildcard).
Comments
Post a Comment