HN user

waldfee

225 karma
Posts0
Comments61
View on HN
No posts found.

i am only dabbling in this space myself, so can't answer everything. all the formats i mentioned are for a quantized version of the original model. basically a lower resolution version, with the associated precision loss. e.g. original model weights are in f16, the gptq version is in int4. a big difference in size but often an acceptable loss of quality. using quants is basically a tradeoff between quality and "can i run it?".

examples of original models are llama(2), mistral, xwin. they are not directly related to any quantized versions. quants are mostly done by third parties (e.g. thebloke[1]).

using a full model for inference requires pretty beefy hardware. most inference on consumer hardware is done with quantized versions for that reason.

[1] https://huggingface.co/TheBloke

i don't think such a guide exists. this space is moving pretty fast. a short rundown

quantized model formats:

- GGML: used with llama.cpp, outdated, support is dropped or will be soon. cpu+gpu inference

- GGUF: "new version" of the GGML file format, used with llama.cpp. cpu+gpu inference. offers 2-8bit quantization

- GPTQ: pure gpu inference, used with AutoGPTQ, exllama, exllamav2, offers only 4 bit quantization

- EXL2: pure gpu inference, used with exllamav2, offers 2-8bit quantization

here[1] is a nice overview of VRAM usage vs perplexity of different quant levels (with the example of a 70b model in exl2 format)

[1] https://old.reddit.com/r/LocalLLaMA/comments/178tzps/updated...

there are modern products in this niche, and there is huge interest. the market certainly seems to be there (regium tried to defraud people for almost a million dollars i believe was the kickstarter sum before they got shut down).

there is squareoff [0], with new products currently in development (swap / neo)

then there was regium, an elaborate scam on kickstarter [1]

now there is phantom [2], which hopefully is not a scam. they at least posted some engineering details on hackaday [3]

squareoff has chess.com support, hopefully with lichess support coming (they are promising it, but has not yet happend). phantom claims working lichess support and to work on chess.com support

[0] https://squareoffnow.com

[1] https://www.chess.com/news/view/update-on-regium-chess

[2] https://www.kickstarter.com/projects/wondersubstance/phantom...

[3] https://hackaday.io/project/179268

OpenDiablo2 6 years ago

OpenRA states in the about section

This means that OpenRA is not restricted by the technical limitations of the original closed-source games: it includes native support for modern operating systems and screen resolutions...

I think this is the most compelling reason - it opens up features that would be way harder to realize in the original engine.

noteworthy

- OpenXcom https://openxcom.org/about/

- OpenRA https://www.openra.net/about/

- OpenTTD https://www.openttd.org/about.html

that's pretty much my train of thought as well.

trickle down economics don't work like the theories would want it to, just look at quantitative easing and the ECB equivalent. all those cheap loans to banks never ended up in the real economy

if you look at the graph, it all begins after 1971. that was the year the gold standard was finally abandoned for real.

i am not an expert and don't know all the important puzzle pieces, or even understand them. but looking at our financial systems, i think a big part of the reasons for the problems of our current form of capitalism is the unlinking of fiat currency from any form of backing.

In my experience, this is not the case when both the reviewer and the reviewee are working together and want to find a good solution.

Rephrasing criticism is a pretty common technique not exclusive to code reviews and imho helps avoid a defensive stance which makes it easier to be productive.

to further clarify the power of qmk: you literally program (parts of the) firmware in c. so you can put modifiers everywhere you want, make layers until you are going crazy, make keys behave differently on press/hold, absolutely everything. and it's firmware you flash, so no matter where you plug it in, it' all there.

qmk is just awesome, check it out https://github.com/qmk/qmk_firmware/

Blizzard also has their own stuff. OATH for everything would make things so much easier. I cannot understand why they wont simply do that.

What i hava a problem with in regards to account recovery options is that some services require you to also have enabled sms based 2FA in addition to TOTP or similiar as a fallback. That defeats the whole purpose of non-gsm based 2FA. The whole construct is as insecure as sms based alone, the TOTP part is entirely useless. at least make it optional if you think some users need it.

I switched to a four day work week (cutting weekly hours, so not working more on those four days) a few months ago. It's glorious. I just use it as an additional weekend day, meaning time with family, hobbies and household chores.

But i am also not a youngster anymore, so side projects and work in general is not my prio #1 anymore