I have found a bug in gen_merged_model.py which i used for model fusion. It is about failure for setting engine to 'CAFFE' when i used this script. I figured out that the return value of layer.convolution_param.kernel_size was always a list rather than a single value. I have fix this and sent pull request.
I have found a bug in gen_merged_model.py which i used for model fusion. It is about failure for setting engine to 'CAFFE' when i used this script. I figured out that the return value of layer.convolution_param.kernel_size was always a list rather than a single value. I have fix this and sent pull request.