Lei Mao

It is not possible to load the frozen model to Keras as far as I know.

https://cors-anywhere.herokuapp.com/ is only intended for local development. It has limits on the number of redirects before re-activation. So it is not a solution for deployment unless you host your own...

应该是[这行](https://github.com/shenliyang/hexo-theme-snippet/blob/a0bae5c6a86f4d6c1b6007001e07a9bc5343f8e5/layout/_widget/social.ejs#L9)不兼容最新的font-awesome >= 5.0。但是我尝试着简单修复了一下没有成功。

@N-Dekker @jwakely I agree with @frederick-vs-ja that the `int32_t i1 = *reinterpret_cast< int32_t * >(&u);` (I changed the `int` to `int32_t` to eliminate the assignment undefined behavior (caused by integer...

How about adding attributes to a custom PyTorch `torch.autograd.Function` operator so that the attributes can be available in the ONNX?

> You may consider using https://github.com/microsoft/onnx-script to convert the model into Python code, edit the attribute, then export the model back to ONNX. It's an early project so not everything...

Please indicate whether you are using the Docker container that I provided. Also please provide your command and all the information printed out in the terminal.

I could not help if the user was not using Docker container because I cannot easily reproduce the bug.

> I agree with this (although I'm not sure about the last bit: why you are mentioning that variance is required for calculating predictive RMSE ....) It has been half...

> @leimao Have you managed to update so the images show up? I compiled your branch but the images still are not showing. Not an expert to change configuration sadly....