Valeriy Fedyunin

Wow. That's the first time I see this problem not as a `CMake Warning` but as an error. I will take a look at that, because I think that should...

This PR changes the interface and that will lead to compilation problems on the user side. That's why we will merge it waaaaaay later (with the next major release). Don't...

@lordnn > Добавление константности к аргументу требует простой перекомпиляции пользовательского кода, а не переписывания Incorrect. In this case the `IClustering` interface was changed. That requires changes in the user code...

> Я не увидел механизма который бы мотивировал пользователя реализовывать свой `IClustering` интерфейс... There are other realizations (not in open-source). Otherwise I would've merged this PR already.

Yep, we know. The work is [in progress](https://github.com/FedyuninV/neoml/tree/PyOnnxExperiments/NeoML/Python/neoml/Onnx) but I can't tell you when it will be moved to master...

Can you please share the implementation of this feature (or at least some paper). Because google shows that this feature is implemented differently. E.g. I've found an implementation (also OHEM)...

Hi there! Thanks for the report. It looks like a bug which leads to infinite recursion. That's why it will be helpful to know how you got the Python package....

Currently NeoML doesn't support training for Vulkan math engine. What remains unclear here: why Vulkan math engine was created instead of CUDA (which is by default when GPU is made...

That will require some time to implement (PS ROI pooling + deformConv for all the platforms, optimizations, etc.), that's why it's not gonna be done in near future (sadly). Feel...

Еще один, возможно, полезный вариант: дополнительная проверка перед скачиванием. Т.е., если файл существует, но другого размера или еще по каким параметрам не подходит - то удаляем тот, что был на...