Lexcess

Ok then I’ll bite. First up, the common refrain is don’t worry Non-Nullable is coming, that’ll sort out 90% of your problems. With fresh eyes I'd like frame method contracts...

I'm not sure the value proposition is there if contracts boil down to moving existing guard code outside the method body. I'd like to see something that is more strict,...

@TylerBrinkley That is a pretty neat solution. I was originally thinking you'd want to separate the predicate from the exception but if we assume that contracts are either applied by...

@MovGP0 interesting, would you not at least agree that a where constraint is a similar concept? I think it would aid discoverability, especially if you are new to the language....

@MovGP0 Yes I use those a lot in the old contracts, but if we are modifying the compiler I'm thinking lets try and get a cleaner implementation here. That model...

@Flash3001 I agree that static checks would be useful. I think trying to do the same amount of work in the compiler as the old static analyzer did is unrealistic...

@TylerBrinkley My thought exactly. @Flash3001 Interesting. I guess my only concern (and this goes for @MovGP0 comments as well) is that I don't think we will get the kind of...

@MovGP0 Apologies, I missed your ref. My gut instinct is that I wouldn't support a ref return constraint. Because ultimately it could be altered from another context so isn't provable....

You were first, but this seems to be the same as #252

As a slight side note to the original premise or your issue, there are far smaller 'difference only' versions of the original controlNet models here: https://civitai.com/models/9868/controlnet-pre-trained-difference-models they are still larger...