Shintarou Okada

@Hakuyume `Sigmoid`, `Transpose`, `Mul`, `Conv(dilation > 1)` are now available on `mkldnn_with_generic_fallback` backend. Please check.

Sorry for my late reply.　Current menoh supports only 1.4.1 . I think FreeBSD package's onnx is not 1.4.1 . So please build. Or, As I told on #226, you...

I see. I'll try to modify.

Hi. I made a branch that is the modified master to be compiled with C++11. You can try it. Please give us little more time to decide to merge it...

Hi. Thank you for your trying. Menoh utilizes [MKL-DNN](https://github.com/intel/mkl-dnn) and MKL-DNN supports SIMD optimization. It is automatically enabled. You needn't to set any option to enable SIMD. Could you tell...

Thank you for your important questions. I'll answer for them. (1) Currently Menoh is an experimental framework and its implementations are incomplete. Of course, there is a design document about...

Thank you for your another important question. NNVM/TVM is good framework and it has partially common goal with Menoh. However they also have differnt perspective about DNN inference. Let me...

Hi @rajh619. Thanks question. Honestly speaking, modifying trained model previously, graph manipulation is not needed much. However, it is troublesome for users who want merely to use trained models distributed...

Hi @kmorito It seems `Constant` operator is contained in imagenet_vgg16.onnx. Menoh 1.2.0 doesn't support `Constant` operator. Now PR for `Constant` is created https://github.com/pfnet-research/menoh/pull/222 . It will be merged soon. Also...

Sorry for my late reply. Flatten node is not supported yet. I estimate Flatten is used to connect the last fc layer in keras's VGG16. Chainer's VGG16 uses Reshape instead...