7
Jedidja
3y

I dun goofed

made a neural net that runs against a simulation. Wanted to run it overnight to get some meanigful stats and insights

But yesterday afternoon I changed something in the simulation and ofc tested it without the nn ... and then forgot to put it back on
So while I expected to come in today and start plotting and analyzing the data while the runs finish, in reality I'm sitting here on a lot of useless data, not knowing what to do.

I kinda want to just start it again and go home

Comments
  • 5
    By my experience, overnight tests SUCK. There is this little djinn that fucks up your data at night and makes exception-and-error prone data only ever be run when you are not looking.
    Can't really escape it, though.
  • 2
    @JsonBoa
    Easy. Run tests in parallel, in different processes, so when they crash and burn, they will continue testing, and not get stuck.
  • 1
    @magicMirror oh, yes, you can workaround exceptions when crunching a dataset, and if exceptions are rare enough it may suffice.
    Shit hits the fan when you lose significant quantities of datapoints to some stupid data format inconsistence or an edge case something. Those things you could easily fix were you there to see it.
  • 1
    aaand now I had a division by zero.
    it ran for an hour after I left the office

    not the most efficient way to find bugs I guess
  • 0
    @JsonBoa

    > There is this little djinn that fucks up your data

    Yes, but does he grant wishes?
  • 1
    @Wisecrack most surely it makes you wish for grants. In no small part because you spent most of your time re-doing failed experiments.
  • 1
    @JsonBoa Well played.

    Could be like me, and spend five years of your life, banging your ahead against a brick wall with basic algebra. All to invent a lousy non-sieve prime generator.
Add Comment