Bump dependency compat and try using TO on newer CUDA/cuTENSOR

[kshyatt]

This actually failed for me locally which is interesting/concerning. It would be great to support newer cuTENSOR and CUDA.jl as a lot of updates have happened there.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[kshyatt]

That makes sense. I created this mostly as a test to see if this would instantly break everything. Where is it that we’re relying on this specific version? Is it in a downstream package?

…

On Thu, Sep 4, 2025 at 12:56 PM Lukas Devos ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In Project.toml <#223 (comment)> : > @@ -49,7 +49,7 @@ StridedViews = "0.3, 0.4" Test = "1" TupleTools = "1.6" VectorInterface = "0.4.1,0.5" -cuTENSOR = "2.1.1" I know it doesn't matter too much for this PR, but I think this minimal version was really required, mostly because we are relying on this PR : JuliaGPU/CUDA.jl#2356 <JuliaGPU/CUDA.jl#2356> — Reply to this email directly, view it on GitHub <#223 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGKJY4TEDGVSTZRKYNGFXL3RALETAVCNFSM6AAAAACFTNUF2KVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTCOBUGU4DIMBTGE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[lkdvos]

No, the cuTENSOR.jl wrappers were restricted to DenseCuArray types before the PR I linked, and we are wrapping everything in CuStridedView = StridedView{<:CuArray}, which is not a subtype of that unfortunately.

[kshyatt]

Ah ok I see, I'll restore the minimum version then

[Jutho]

@kshyatt , completely forgot about this one. I assume this is still good to be merged as is?

[kshyatt]

Let's resolve the conflict and make sure tests still pass, but then yes, should be good.