Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)T
Posts
0
Comments
154
Joined
2 yr. ago

  • Really? What do you expect is the edit rate on sites like Lemmy and reddit? One in ten comments? I think more like one in 30 or something. That would increase the storage costs by 3% and a small amount of processing power.

    Hosting costs are dwarfed by media storage anyway.

  • Are updates authenticated? Or can I send an update to lemmy.world from 123.123.123.123 (which is not the IP address of feddit.de) that you have edited your comment to say "I don't like pizza"?

    If updates are not authenticated this really could be a big problem.

  • Knowing how comments get changed is immensely interesting data. And if you design a system from the ground up, adding the functionality to save edits in the backend does not take much effort at all.

  • I have a graphical application that crashes regularly when I switch between displays with Ctrl+alt+number. Something in the winit stack does not like it.

  • Compilation: top row, runtime: button row.

  • Run an open source one. Training requires lots of knowledge and even more hardware resources/time. Fine tuned models are available for free online, there is not much use in training it yourself.

    Options are

    https://github.com/oobabooga/text-generation-webui

    https://github.com/Mozilla-Ocho/llamafile

    https://github.com/ggerganov/llama.cpp

    I recommend llavafiles, as this is the easiest option to run. The GitHub has all the stuff you need in the "quick start" section.

    Though the default is a bit restricted on windows. Since the llavafiles are bundling the LLM weights with the executable and Windows has a 4GB limit on executables you're restricted to very small models. Workarounds are available though!

  • Why the low resolution image though? Or do you really have 1366p screens and like 5px high fonts?

    WM looks sexy btw

  • There are 4x the pixels so.....

    Totally agree, but then

    2k, 4k, and 8k

    Is internally inconsistent!

    If 4k is four times the pixel count of 1080, then 2k means 1440 (-ish, it should be 1530) - that's fine. But then 8k must be 3050, but it is actually 4320!!!

    So it can not refer to the number of pixels (quadratic scaling). On the other hand, if we assume linear scaling and 8k is 4320 and 4k is 2160, then 2k is 1080 - but 2k is never used in that context!

    Edit: as you can see I'm very passionate about this XD

  • Me running an LLM at home:

    The same image, but the farmer is standing in front of a field of poppy (for opioid production)

  • Maybe consider paperless-ngx.

    Its primary purpose is document management, but you can easily upload receipts and pictures as well. I use paperless-mobile to interact with my instance.

  • I have the opposite problem, llavafiles (a large language model, packages as a single files) can run on both Linux and Windows. They are written to be compatible with both.

    But when I ./file to run it, eine is started automatically!

    (The llava file GitHub has a workaround, but still by default it chooses wine for some reason)

  • It's a more easily readable drop in for A and B. And it being convention helps remove one unfamiliar element from a new topic.

  • I can see where you are coming from.

    But my impression is that Japan has a lot of unspoken rules on how to interact with others. And unspoken social rules are the arch enemy of autism, lmao.

  • Define "it"

    Because waifu stickers may indeed speed up "it" for some definition of "it"

  • The prequels are already leaked on Reddit and Lemmy.

    !I don't recommend it, but if you want to, you can watch a game of tag between an explosive drone and a human. Spoiler: the drone won.!<

  • We already have AI in Firefox. And not gonna lie, offline (I.e. absolutely private) translations for webpages is pretty neat.

  • Ideally they don't need actual accidents to find errors, but discover said issues in QA and automated testing. Not hitting anything sounds like a manageable goal to be honest.

  • And yet people enjoy it.

    Maybe we should build our city centers the same way then.

  • No they don't. They say there is a difference between Smartphones and newspapers or books.