Iaroslav Postovalov

😅 Duplicated by #472

@Ixw123 you now can use `DoubleTensorAlgebra`, also SVD is supported in `kmath-ejml`: ```kotlin import space.kscience.kmath.ejml.EjmlLinearSpaceDDRM import space.kscience.kmath.linear.SingularValueDecompositionFeature import space.kscience.kmath.linear.getFeature import space.kscience.kmath.linear.invoke import space.kscience.kmath.linear.one EjmlLinearSpaceDDRM { val svd = getFeature(one(1, 2))!!...

Blocked by https://youtrack.jetbrains.com/issue/KT-46899 because I don't see a concise way to share test cases code.

> Hey, this is a very important feature for my project. Is there any chance of upstream support? Same

> OK, I will try, but we need to check if it will affect performance for large geometries. Maybe there just should be an option to disable it?

Seems to be related to https://github.com/Kotlin/dokka/issues/2024

I quite agree with you about Gradle... I personally understand around 30 classes from the Gradle core.

Also, quantization even to 4 bits may be possible, like it is successfully done for LLaMa. https://github.com/ggerganov/llama.cpp

Version for 1.8.20 is ready but not yet published (cc @nbirillo ). Anyway, you can also consider if https://github.com/google/ksp, which is maintained with larger resources, covers your use case.