Anton
Anton
The PR helped to troubleshoot an issue in the internal CI. I'm closing it, since the issue has been resolved.
@icfaust, we are targeting to add the support in `0.17.0` release of dpnp.
#1785 covers that PR
The numbers with _dpnp=0.12.0_: | example | numpy | dpnp CPU | dpnp GPU | size | | :--- | :---: | :---: | :---: | :---: | | game...
The numbers with _dpnp=0.12.0_: | workload | numpy | dpnp CPU | dpnp GPU | size | | :--- | :---: | :---: | :---: | :---: | | gpairs...
This is because in case of `a` as scalar or zero dimensional array, dpnp will fallback on numpy implementation and so the return type will belong to numpy. Btw this...
@stisa, thank you for noticing that. As per documentation `dpnp.repeat` supports input array `x` of either `dpnp.ndarray` or `usm_ndarray` type. We will prepare a PR to raise an explicit `TypeError`...
`dpnp.logaddexp` is added by #1561
`dpnp.rint` is added by #1537
`dpnp.signbit` is added by #1535