Clément Pinard

Got the same error here. ``` Ubuntu 16.04 Cuda 10.0 Pytorch 1.1.0a0+7e73783 (built from source) python 3.7 ``` although solution from #21 seems to work. Discussion from #15 also hints...

After some investigations, it seems related to gcc version. Originally tested it in gcc-7 but it didn't work. Changed to gcc-5 with a simple "update alternatives" and now it works....

Hi thanks for the commit ! unfortunately, I believe the `fminf`and `fmaxf` is implicitely casting everything to float32. As a consequence, the `check.py` and `grad_check.py` are now broken with cuda,...

I'm pretty sure you need to specify a size for the shared array `sdata` Here if I understood correctly it's 2* 512 ? You should probably define these numbers at...

Ah actually you are right, you can use dynamic shared arrays. Apparently, you need to specify the shared object size in another option in the kernel call. See here :...

Ok sorry about misleading you, your code is mostly fine. I tried your code, and the problem seems to come from the template and the fact you use three different...

what are your OS, python, pytorch, CUDA and GCC versions ? what is the full build log ?

The K80 enabled servers in google colab are not very fast and their architecture (Kepler) is not as easily usable more recent ones like Turing or Pascal You can at...

Hello, I implemented the CUDA/CPU flexibility myself in this project : https://github.com/ClementPinard/Pytorch-Correlation-extension

For your cuda version problem : You most likely have both cuda 10 and cuda 9 installed. What it the output of `nvcc --version`? See [here](https://github.com/pytorch/pytorch/blob/master/torch/utils/cpp_extension.py#L26) about how the pytorch...