Masahiro Sakai

FYI: I encountered another issue that causes the error (#79). One way to solve the issue is checking the exit code of `protoc` and aborting `cmake` when failure. But it...

@Hakuyume and @okapies, #80 solved the problem?

Just FYI: Glow adopted fixed shape of tensors. [Glow: Graph Lowering Compiler Techniques for Neural Networks](https://arxiv.org/abs/1805.00907) > We have considered the idea of developing some kind of parametric tensor types...

Sorry, I overlooked the review request...

It generally looks okay to me. Some notes: * It might be better to package `make install/strip`ed version of libraries/executables, but I'm not sure as I'm not very familiar with...

Note that now there are multiple jobs for same platform due to #81 and #83. It is necessary not to upload same file from multiple jobs.

I dumped the model using following script with modified version of onnx-chainer https://github.com/msakai/onnx-chainer/tree/94baaf4294c7e8ae1a255cc5823535fe1e1dd9e7 : ```python import numpy as np import chainer import chainer_chemistry from chainer_chemistry.models import GGNN import onnx_chainer mb...

With #90 it always fails with `menoh unsupported operator error: Identity`.

@mstksg friendly ping

CI fails with `Out of memory` on windows-i386. The message is printed at https://github.com/ghc/ghc/blob/31cd867e4d37072c3ce1d51efadc94e66ddc1c28/rts/win32/OSMem.c#L77 in `allocNew` function after the invocation `VirtualAlloc(0,0x200000, MEM_RESERVE, PAGE_READWRITE)` failed. My hypothesis: 1. A large number...