syurkevi

Are you using the cuda backend? If so you may also need to add the cuda paths. For example: export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/cuda/nvvm/lib64/:/opt/arrayfire/lib64/:$LD_LIBRARY_PATH More details here: https://github.com/arrayfire/arrayfire-python/wiki If that doesn't help, can...

This is a mistake in the dev branch, the master branch should still have the correct code. Thank you for catching that! I'll make sure to address if we decide...

https://github.com/arrayfire/arrayfire-ml/pull/36#issuecomment-317360576 ArrayFire has lookup, however its behavior seems a bit weird, would it need to be changed in ArrayFire or just emulated in autograd?

I was only able to reproduce using the docker. Thank you for that! The issue seems to be the double linking of two different OpenMPs, no good! Adding -Wl,--as-needed to...

I haven't had a chance to look at this closely yet, but I did a bit off digging in the fitpack code. It seems that fitpack is using B-splines whereas...

Addressed by #3357. Related to issues #2214, #2215, #2244. Correct resolution and release version tbd.

I've found a few kernels this week that fail for larger sizes, I think this may be related to those failures as well. I will re-open the issue if I'm...

Yes, this issue is in the core repo in one of the where kernels [here](https://github.com/syurkevi/arrayfire/blob/e7ffdd080eae16b2d517a5e41feab5c39a748bc1/src/backend/cuda/kernel/where.hpp#LL63C31-L63C34) as well as in the scan kernels. This is a larger issue due to the...

I think that assessment is accurate. Do double check your tool is only targeting the CPU backend. We use standard C++ code and porting the CPU backend shouldn't be too...