NeedsMoar
NeedsMoar
TL;DR - It's probably never worth using TTA. UHD is broken. TTA is some kind of undocumented filtering that's normally done (on GPU) by: 1. Creating 8 extra buffers to...
No, the code is just broken. When processing an entire directory, it always uses numframes / the count of files to control this. You need to specify how many frames...
There are conditioning nodes that parse this out too, although I have no idea how they're supposed to work with no model output unless that's implicit which is kinda worse...
Sorry, missed this. I'll whip one up for you soonish, shouldn't take long since that's pretty much all of it up there.
Did anyone else happen to notice the more important part there... Torch's limit for the GPU is 16 **_EXABYTES_**? 17179869184.00 GiB I can only guess there's a signed / unsigned...
The DLLImport thing is sort of a red herring, it's only there to generate headers for things that link to the DLL and need to know about the class interfaces...
The loaders don't take much IO time at all with non-loaded models (none if the checkpoint is still in standby / ram cache which is a function of the OS...
I can't recall what square brackets were supposed to do... were they implicit downweighting like how () upweights slightly?
According to the community manual [] doesn't do anything in prompts. Downweighting is via () with a number less than 1 specified. [] did downweighting in shark, so the equivalent...
I just did a once-over on the conditioning parser. [] do nothing.