Problems on ReservoirComputing.jl

I’m now trying to use ReservoirComputing library, but when I try to set the NLAtype, it doesn’t work and shows “UndefVarError: NLAT2 not defined”.
I wonder what’s wrong with the code
(You can find the code here:ReservoirComputing.jl/rossler_example.jl at master · SciML/ReservoirComputing.jl · GitHub)

What version are you using?

julia 1.7.1

No, ]st ReservoirComputing

I download Julia 1.6.2 and it works, thanks a lot !

Sorry to but in but this seems suboptimal - what Chris meant was that you were likely using an outdated version of ReservoirComputing when you saw the error, which is why he asked for the output of ] st.

You now downloaded a different Julia version which uses a fresh global environment, so that when you installed ReservoirComputing you likely got the latest version and things are working. This is not because of any issues with Julia 1.7 though, but likely because you had other packages installed in your global environment which held back ReservoirComputing to an older version.

So there’s two takeaways here I think:

  1. Work in project specific environments and don’t install packages in the global environment unless they are development packages like Revise, Pluto etc.

  2. Use the latest stable release of Julia for your work - this is currently 1.7.2 and will soon be 1.8. There’s no reason to use 1.6.2 and it is extremely unlikely that there are any well maintained packages which work on an older version but not the latest release

v1.6 is fine though as it’s the LTS

That’s a fair point although I’d say (1) the current LTS release is 1.6.5 not 1.6.2 and (2) as per many previous discussions almost everyone should be using stable rather than LTS

Yeah, everyone should probably be using stable, it’s suboptimal not to be, and this case will work on v1.7 if the person just updated their packages. But, LTS is maintained by SciML so this should work until v1.6 is not the LTS. That’s part of our development plans, and it is wrong to say “there’s no reason to use 1.6.2 and it is extremely unlikely that there are any well maintained packages which work on an older version but not the latest release”. It is extremely likely the SciML packages do as LTS is CI tested, and please open an issue if an unknown issue is found.

Yes, my point was about “older version” in general, not LTS in particular - still I’d argue that it is unlikely that you’ll find a package that works on LTS but not on stable. Currently of course we only have one post-LTS minor release so my point about “work[ing] on an older version but not the latest release” isn’t all that relevant, but once 1.8 is out I’d say it is indeed quite unlikely to encounter a package that works on 1.7 (and LTS) but not on 1.8.

Got it. thanks a lot

1 Like