I noticed that CPU-only mode was removed at from this fork of Caffe. Why? I can definitely see a use case for not wanting to compile the entire CUDA-enabled library for quick testing (especially on an embedded device) before compiling the full CUDA-enabled pipeline.
I noticed that CPU-only mode was removed at from this fork of Caffe. Why? I can definitely see a use case for not wanting to compile the entire CUDA-enabled library for quick testing (especially on an embedded device) before compiling the full CUDA-enabled pipeline.