Xiu-zhe (Roger) Luo
Xiu-zhe (Roger) Luo
> I personally guessed runtime type-inference yields large runtime overhead and thus language like julia might have disadvantage when overhead becomes important. So it is interesting for me that this...
No, you need to dispatch them to specific rules as well. E.g ij,jk->ik should be dispatched to inplace mat mul, e.g mul!
Should be straight forward to support multivariate Hessian with the new grad.
yeah, next week maybe, after I finish my assignments.
also #50 I think. then it should cover most of the use cases.
I think this is a great idea. I had a tomography framework (in private, a draft) for GAN based tomography with pytorch last year (GAN is failed on the tomography...
Here is some of my legacy code for QST & VMC for reference: Pure python implementation: https://github.com/Roger-luo/legacy-QMTK With C++ optimized: https://github.com/Roger-luo/legacy-QMTK/tree/cpp Some functionalities in the package was separated to other...
yes, @ChrisRackauckas that works! I'm not sure what's the difference with or without AD tho?
I currently manually workaround this by dividing this triple loop using https://github.com/QuantumBFS/BQCESubroutine.jl/blob/multithreads/src/schedule.jl ```julia function broutine2x2!(S::AbstractMatrix{Complex{T}}, U::AbstractMatrix, locs::Locations) where T plain_locs = plain(locs)::Int step_1 = 1 >plain_locs)-1) @inbounds @batch for tid...
是的，有些关键字的文档还没有。