sugarme

@lieral , From error logs (the initial code I read from my email), it seems that some tensor was deleted 2 times (illegal). However, I couldn't see it here now...

@lieral , I can reproduce the issue in a for loop. Forward loop seems to be unstable and might related to memory allocation and handling from Go to C. I...

@lieral , Would it be able to share your Pytorch model code (Python?). I would like to see how output tensors are formed and returned.

@Kur-Ich , We haven't got a setup instruction for Windows yet! There's different `libtorch` for Windows. You can download from Pytorch site: - [Libtorch CPU](https://download.pytorch.org/libtorch/cpu/libtorch-win-shared-with-deps-1.10.0%2Bcpu.zip) - [Libtorch CUDA](https://download.pytorch.org/libtorch/cu111/libtorch-win-shared-with-deps-1.9.1%2Bcu111.zip) and unzip...

@Kur-Ich , @chnykn , I now can setup a Windows box for testing. Although the issue you mentioned above related to linking issue, it can be fixed by wiring up...

@ZonghaoWang Can you give more detail - code example and how you measure your performance please?

Your time measurement was taken for other steps after ts.TensorFrom. Can you just break down just calculate time for ‘ts.TensorFrom’ only? And how do you judge it is slow? In...

Any how slow to compare to Python Pytorch? Or other framework?

Thanks @rcoreilly for the suggestion. Actually, I was struggling with such naming and still. I chose prefix `Must` for 2 things: 1. It reminds keep in mind that our program...

Yeah, quite "musty" :). However, some may argue that the better way of using APIs is return "error" instead of "panic" so that you have a chance to handle error....