Seth Axen
Seth Axen
> harder to wrap the results to be accessed in the same way. What do you mean by this?
> The output of Optim has a specific format. I'm not sure that it is possible to construct the complete "results" type from what we get from LBFGSB and the...
> Oh, okay. Thank you for all the work you did looking into it. No problem! Was actually pretty straightforward. > Okay, I'm thinking if it should I've in Optim...
@pkofod Any further thoughts on this? If we keep OptimLBFGSB.jl, then I would finish the tests and register it.
I'll go ahead and write the tests and register OptimLBFGSB.jl.
I'm seeing this as well. The results of the last 2 lines in the above example are on my machine: ```julia julia> @show norm(results_single .- results_thread_v1); norm(results_single .- results_thread_v1) =...
> The `cfg` object is used as a cache / work buffer so sharing it between threads is not possible. It seems then that `cfg` is only safe to use...
Was this completed by #2510?
Also @andrjohns is this the last major prerequisite for implementing the Gamma quantile and its derivative?
Agreed, this should definitely be fixed. Probably as simple as using a hermitian projection (which is I think a utility function elsewhere in ChainRules) instead of using `triu!`.