Made sample type agnostic and GPU compatible
#93
Conversation
| @@ -655,6 +655,8 @@ function sample(rng::AbstractRNG, m::MPS) | |||
| error("sample: MPS is not normalized, norm=$(norm(m[1]))") | |||
| end | |||
|
|
|||
| ElT = promote_itensor_eltype(m) | |||
There was a problem hiding this comment.
I've been preferring scalartype(m) for this these days, probably we will remove promote_itensor_eltype.
mtfishman
changed the title
sample type agnostic and GPU compatible
|
Thanks for the quick PR! This is subtle to test right now, we have to set up the tests in this repository to run on GPUs, I guess we'll address that in future PRs. |
|
@mtfishman I can look around other code in |
|
Let me know if you want some help with that @kmp5VT. I'd be willing to spend some time on that. |
|
Thanks @kmp5VT. Also it would be great to start switching some tests over to being able to test on GPU, looping over device backends the same way we do in the NDTensors/ITensors tests. For now it could just test using JLArrays, but ultimately we should also set things up with Jenkins to run the tests on GPUs. |
Fixed the issue of
samplementioned in #92.Closes #92.