Jutho

Sure. I don't think it should be registered as we probably want to reorganise some stuff. It contains TensorOperations specific stuff that is not really TensorKit. I want to invest...

So far, it doesn't. It still depends on a call to `similar` for one of the arguments. However, the redesigned code certainly makes it easier to change this behavior, and...

Thanks for this question. I don't think there is a well established reason for preferring an absolute tolerance as I currently do. The only problem is indeed with eigenvalues close...

I am still not fully convinced. I prefer that the algorithm uses some absolute tolerance. If you want something that scales with the overall scaling of `A`, then the absolute...

I don't understand why you use inverse iterations for the minimum eigenvalue of a Hermitian positive definite operator? It's not because it is close to zero, that you need the...

A possible solution could be to start using the API provided by ArrayInterface.jl

I have not had any time for coding / package maintenance the last few weeks. I will try to find some time over the next couple of days.

Hi @albertomercurio , my apologies for the late response. I think your solution using `AbstractGPUArray` is currently the best one. However, I removed Manifest.toml from the PR. (I actually did...

> Ok I should fixed it by introducing the AbstractGPUArray type from GPUArrays.jl > > However, the first time I run the solver with the GPU it returns the out-of-memory...

Can I infer correctly that you have 64 GB of memory on your machine? But you still use a 32-bit version of Julia. How does that work exactly?