I have, they degrade very significantly whenever the context goes to 8000+ tokens.

I think there are graph/tree based solutions for rag, which ideally should have the books at leaf nodes, and their overlapping summaries as parent nodes.

Kinda like BST for books to rag from.

It’s a big problem with reddit like forums where the comments are highlighted and much more densely presented than the article itself.

Older forums avoided this (unintentionally) by not having any sorting on replies and keeping the density of replies, low.

Well okay, I’ll be skipping this then.

Problem isn’t the meat industry, problem is that it is prohibitively expensive

That’s the best gpu for this usecase. Maybe use 2x if you want the best of the best.

Is there a way to install it on baremetal windows? I’m kinda tired looking for a document hostler that can work on baremetal windows atp. Everything either has docker method or doesn’t support windows.

ILibrarian is the only one offering that.

Lmao. Lemmy’s active users isn’t even 1% of people pirating.