fhein

Lite fredagsunderhållning åt folket

Couldn't find a dedicated community for distro recommendations, I hope it's ok to ask here.

A couple of years ago my wife and I built a computer and gave it to a friend's kid. We put ElementaryOS on it since that seemed pretty fool-proof, but it appears to require a re-install to upgrade major versions so it has been stuck with an old glibc and because of that he can't play Factorio.

For his 13:th birthday we bought him a SSD so it would be a good time to reinstall Linux, but is there perhaps some better choice than ElementaryOS? They live quite far away so I can't easily pop over to fix his computer if something breaks, we don't spend enough time there for me to teach him to fix things himself, and he doesn't seem very interested in learning how computers/operatings systems work either.

• Hardware: Some old Intel CPU with 8GB DDR3 and a GTX1080
• Usage: Gaming through Steam+Proton, Lutris and browsing.
• Requirements: Games work, OS never breaks on updates. Doesn't need to be "kid proof", I don't think he touches any stuff he doesn't know what it does.

In case anyone isn't familiar with llama.cpp and GGUF, basically it allows you to load part of the model to regular RAM if you can't fit all of it in VRAM, and then it splits the inference work between CPU and GPU. It is of course significantly slower than running a model entirely on GPU, but depending on your use case it might be acceptable if you want to run larger models locally.

However, since you can no longer use the "pick the largest quantization that fits in memory" logic, there are more choices to make when choosing which file to download. For example I have 24GB VRAM, so if I want to run a 70B model I could either use a Q4_K_S quant and perhaps fit 40/80 layers in VRAM, or a Q3_K_S quant and maybe fit 60 layers instead, but how will it affect speed and text quality? Then there are of course IQ quants, which are supposedly higher quality than a similar size Q quant, but possibly a little slower.

In addition to the quantization choice, there are additional flags which affect memory usage. For example I can opt to not offload the KQV cache, which would slow down inference, but perhaps it's a net gain if I can offload more model layers instead? And I can save some RAM/VRAM by using a quantized cache, probably with some quality loss, but I could use the savings to load a larger quant and perhaps that would offset it.

Was just wondering if someone has already done experiments/benchmarks in this area, did not find any exact comparisons on search engines. Planning to do some benchmarks myself but not sure when I have time.

Just posting this since I spent over an hour trying to figure out why I couldn't open my desktop today.. After booting and logging in I got a black screen. Switched to a terminal but did not see any obvious errors in the logs.

Not fixed for Fedora 41 KDE yet, so I installed plasma-workspace-x11 to use in the meanwhile. Anyone who hasn't updated to 6.3 yet could probably change their display settings to not use ICC profiles to avoid it.

Going through some boxes and found a stack of old White Dwarf. I'll keep the first issue I ever bought as a memory but planning to get rid of the rest. Just wanted to check if there are people collecting these before they go into the recycling bin. If anyone's interested I can make a list of which ones I have, and I'll send them to anyone willing to pay for postage. Located in Sweden.

I have calibrated my monitors to create icc profiles, they show up in KDE color management and everything used to work exactly as it should. Now every time I start my computer it goes like this:

1. I log in to my account
2. It shows my desktop, with the right colour correction.
3. After a few seconds the colours revert to look un-calibrated on both monitors.
4. I restart the colord service and it loads the colour correction again.

As an alternative to step 4, if I go to KDE colour settings, select the default profile and then back to my profile then it also starts looking good again.

This problem must've started a week or two ago, but unfortunately I haven't been able to pinpoint exactly when. I haven't touched anything related to colour management in months, and don't think I've done any changes to my system other than upgrading packages.

Can't see anything colour related in the syslog except colord loading the correct profiles. I removed all the old profiles that I wasn't using anyway. I removed dispcal's profile loader from autostart to make sure it wasn't interfering with something. The profiles are both installed system wide and in my user folder.

Using Fedora 39 KDE.

Anyone have any idea what could be wrong, or even how to debug this?

I'm trying to learn more about LLMs, but I haven't found any explanation for what determines which prompt template format a model requires.

For example meta-llama's llama-2 requires this format:

...INST and <> tags, BOS and EOS tokens...

But if I instead download's TheBloke's version of llama-2 the prompt template should instead be:

SYSTEM: ...

USER: {prompt}

ASSISTANT:

I thought this would have been determined how the original training data was formatted, but afaik TheBloke only converted the llama-2 models from one format to another. Looking at the documentation for the GGML format I don't see anything related to the prompt being embedded in the model file.

Anyone who understands this stuff who could point me in the right direction?

Maybe I'm using the wrong terms, but what I'm wondering is if people are running services at home that they've made accessible from the internet. I.e. not open to the public, only so that they can use their own services from anywhere.

I'm paranoid a f when it comes to our home server, and even as a fairly experienced Linux user and programmer I don't trust myself when it comes to computer security. However, it would be very convenient if my wife and I could access our self-hosted services when away from home. Or perhaps even make an album public and share a link with a few friends (e.g. Nextcloud, but I haven't set that up yet).

Currently all our services run in docker containers, with separate user accounts, but I wouldn't trust that to be 100% safe. Is there some kind of idiot proof way to expose one of the services to the internet without risking the integrity of the whole server in case it somehow gets compromised?

How are the rest of you reasoning about security? Renting a VPS for anything exposed? Using some kind of VPN to connect your phones to home network? Would you trust something like Nextcloud over HTTPS to never get hacked?

The only Linux questions community I found appears to be locked, so I hope it's ok to ask here..

For a very long time I've had the issue that occasionally, perhaps 1 out of 40 boots, my mouse does not work once the OS starts. The mouse appears to turn on during POST/BIOS/GRUB, then it goes dark again while the OS is booting, and to make it turn back on I have to crawl under the desk, unplug it and plug it back in. 39 times out of 40 (approximately) it just goes dark briefly and turns on in time for the OS. The mouse also works just fine in UEFI.

This happens in Xubuntu, Fedora and Windows, and with two different mice from different manufacturers. I've also upgraded the motherboard twice and pretty much every component in the computer. The only thing that's always has been there is GRUB, which is why I suspect it could be involved.

It happens so rarely that I never really bothered to try to find a solution for it, but it is kinda annoying when it happens.

cross-posted from: https://lemmy.world/post/1396665

I'm about to invest in some "prosumer" level 18V power tools, e.g. Makita, Dewalt, Metabo, etc. The general consensus appears to be that they're all more or less equal and it doesn't matter much which brand one goes with, but a few years ago I developed tinnitus which has made me more sensitive to high frequency sounds. And having tested a few cordless drill drivers at various hardware stores I can definitely tell they're not equal when it comes to what the noise sounds like, even if they measure similarly in sound pressure level.

Currently I have a Ryobi 18V brushless that I measure to 83-84 dB(c) at 1m which makes it pretty loud, and it also has a quite a quite high pitched sound which makes it worse. Looking at a spectrogram from a calibrated measurement mic there's a pretty clear peak at 6kHz. The drill drivers I've been looking at all have a rated sound pressure level (Lp) around 76 dB so they should already be noticeably less loud, but I'm hoping to find one without those high frequencies if possible.

I've been able to test a few drills in person and got some initial impressions. Unfortunately the local hardware stores have a pretty limited selection of brands, they don't have all models in the store, and you usually have to bother the staff if you want to try something with batteries in it. At first I thought I could remember my subjective opinion of different drivers, but when I later got the opportunity to re-compare two of them head to head I realised I have really bad memory..

Head-to-head comparisons that I've done:

Makita DDF484 vs Dewalt DCD791: The Makita has a very high frequency whine, kind of like dentist's drill. The Dewalt is probably just as loud, but a little lower pitch making the noise slightly more tolerable, while still not great.

Makita DDF484 vs 485 vs 486: When it comes to high pitched noise, 484 was worst, followed by 485, and the 486 was most tolerable.

Dewalt DCD791 vs DC800: The 800 had a more high pitched noise than the 791, once again reminding me of the dentist's drill.

Not compared with anything, so only subjective impression:

Hikoki (a.k.a. Metabo HPT in USA) DV18DD and Dewalt DCD777: Smaller and weaker tools but much more quiet than everything else. If I were going to buy 2 drill drivers I'd probably get something like these plus one larger for heavier work.

Metabo (the made in Germany, non-HPT, brand) DS 18 LT BL: Only had a quick try with this machine but without having anything to compare it with I thought it sounded pretty reasonably tolerable. There's a version with quick change chucks, unfortunately only the smaller (L) and larger (LTX) models have offset and angle chucks for some reason but that's a feature I would really like to have.

Anyone compared Milwaukee to other brands with respect to noise? The only store I found that sells them said they didn't have any charged batteries so I couldn't try any.

Not exactly sure what I'm expecting from this post, since the question is so subjective.. Perhaps someone with experience of multiple cordless drill drivers could share their opinion? Or someone else with a similar aversion to high pitched noises could say if they found any tool that they're happy with?

I'm about to invest in some "prosumer" level 18V power tools, e.g. Makita, Dewalt, Metabo, etc. The general consensus appears to be that they're all more or less equal and it doesn't matter much which brand one goes with, but a few years ago I developed tinnitus which has made me more sensitive to high frequency sounds. And having tested a few cordless drill drivers at various hardware stores I can definitely tell they're not equal when it comes to what the noise sounds like, even if they measure similarly in sound pressure level.

Currently I have a Ryobi 18V brushless that I measure to 83-84 dB(c) at 1m which makes it pretty loud, and it also has a quite a quite high pitched sound which makes it worse. Looking at a spectrogram from a calibrated measurement mic there's a pretty clear peak at 6kHz. The drill drivers I've been looking at all have a rated sound pressure level (Lp) around 76 dB so they should already be noticeably less loud, but I'm hoping to find one without those high frequencies if possible.

I've been able to test a few drills in person and got some initial impressions. Unfortunately the local hardware stores have a pretty limited selection of brands, they don't have all models in the store, and you usually have to bother the staff if you want to try something with batteries in it. At first I thought I could remember my subjective opinion of different drivers, but when I later got the opportunity to re-compare two of them head to head I realised I have really bad memory..

Head-to-head comparisons that I've done:

Makita DDF484 vs Dewalt DCD791: The Makita has a very high frequency whine, kind of like dentist's drill. The Dewalt is probably just as loud, but a little lower pitch making the noise slightly more tolerable, while still not great.

Makita DDF484 vs 485 vs 486: When it comes to high pitched noise, 484 was worst, followed by 485, and the 486 was most tolerable.

Dewalt DCD791 vs DC800: The 800 had a more high pitched noise than the 791, once again reminding me of the dentist's drill.

Not compared with anything, so only subjective impression:

Hikoki (a.k.a. Metabo HPT in USA) DV18DD and Dewalt DCD777: Smaller and weaker tools but much more quiet than everything else. If I were going to buy 2 drill drivers I'd probably get something like these plus one larger for heavier work.

Metabo (the made in Germany, non-HPT, brand) DS 18 LT BL: Only had a quick try with this machine but without having anything to compare it with I thought it sounded pretty reasonably tolerable. There's a version with quick change chucks, unfortunately only the smaller (L) and larger (LTX) models have offset and angle chucks for some reason but that's a feature I would really like to have.

Anyone compared Milwaukee to other brands with respect to noise? The only store I found that sells them said they didn't have any charged batteries so I couldn't try any.

Not exactly sure what I'm expecting from this post, since the question is so subjective.. Perhaps someone with experience of multiple cordless drill drivers could share their opinion? Or someone else with a similar aversion to high pitched noises could say if they found any tool that they're happy with?

I've been toying with the idea of replacing all my old, mostly corded, power tools with a single 18V system and like many others I'm trying to decide between Makita and Dewalt. They seem to be pretty even and most recommendations boil down to "pick the colour you like best".

However I bought a subscription to a magazine that does a lot of product testing, and saw that they gave all Makita drills the lowest rating in the "expected battery life time" category because they only lasted 300 charging cycles out of the 400 they do as part of the test. Now 300 cycles is quite a lot for a home DIY:er, but is battery life something that Makita owners have had issues with? I can't remember seeing anyone complaining about it when looking for user reviews and comments. However I do have a vague memory of some pros complaining about Makita "smart" tools shutting down supposedly because they detect that something might break, thus forcing the owner to hand it in for service.

Article in Swedish, and probably paywalled.

edit: Since I doing a bit of testing how annoying (i.e. mostly how high pitched) different drill drivers sound, which hopefully could be useful to someone else, I've updated the title to include this.