Arseny Tolmachev
Arseny Tolmachev
@lash-1997 0.6.0ではこの問題は再現できるでしょうか？
Actually, it should work with the version in `develop` branch, however the functionality is not tested and will not be officially supported.
You need not simply a dataset, but a segmentation dictionary and annotated corpus. We need to release our segmentation dictionary, but there is almost no documentation on how to use...
Is there a reason why you can't use 0.6.* version?
I am using the same procedure (the only difference is that I was using mingw-msys64 cmake and not the Windows installer) and it works (with AppVeyor CI as well). I'll...
Nice to know that the build have passed. That means that at least MSBuild files generated by CMake are sane. I wonder that is the problem with the VS Solution....
I have looked over your build log once more. It seems that by default CMake tries to generate build files for 32-bit. I haven't tested that configuration. It probably should...
Win32 build should not fail. It is a bug. It should be significantly slower (my guess is ~10x) than Win64 version though.
Tests still fail, so I reopen this issue. Still, I don't know if this is a good idea to support 32 bit architecture at all. The problem is: Juman++ memory...
That's a bug, it should use uint64_t. Otherwise models will not be portable.