>>106530166
Write in the city where the specific office of the company they want you to work at is located

here...

>>106530166
Online

>>106530166
Eritrea

>>106530166
Yes, and don't call me Where

File: 1738162708027.png (940 KB, 616x925)

940 KB PNG

lmao when did civitai get based

This any good?
https://bananaai.live/

>>106531463
1) not a local model
2) if you want to use nano banana, go for google ai studio instead
https://aistudio.google.com/prompts/new_chat?model=gemini-2.5-flash-image-preview

>>106529582
what objective criteria do you use to determine "better" vs "worse"
non-fucked up hands is probably a decent metric
but part that a lot of it is just stylistic fashion
a lot of the older gen pics that looked "great" back then are objectionable now because the style looks so generic and overplayed
people are ever searching new highs of uniqueness and the ephemerality of ai will just make it all the more pathological
nothing really means anything and explicitly exploits that existential vapor part of the psyche to a merciless degree, some percentage of people are going to crash out bad if they don't realize this early on

File: based.png (763 KB, 640x1017)

763 KB PNG

>>106531380
>94.5 gb of vram
VRAMgod sama!

Friend has a PC whose motherboard was 'shorted out' according to the PC shop he sent it to. It just wouldn't turn on one day.
They replaced the motherboard, but now he's getting crashes to desktop and no display issues when gaming. He's getting around it via hardware tweaks (undervolting his GPU and getting rid of daisy chain PSU power cables for two separate ones for the GPU) but the issues still come back intermittently.
I asked him if the shop did a clean reinstall of Windows after they replaced the motherboard, and apparently they did not. Now I'm wondering if Windows driver conflicts are causing his issues.

>>106531116
>I was hoping the HP recovery partition would have no impact in the Win10 install. These couple of posts give me the confidence I wanted to to assume deleting the partition wouldn't need any other cleanup Thanks for the advices & resources.
Nah, it's not tightly connected to the installed OS; after you delete the partition, the next time you try to access "System/Factory Reset" or something like that from the firmware boot menu, it might whine that there's (obv.) no OEM recovery partition located
But let's be honest, the existence of it is "ease of convenience" rather than "very critical don't delete or you will brick your hardware"
>Just wanted to know if GPT would give any stability or performance enhancements that would be worth it.
For starters, you will overcome MBR's limitations like the maximum limit of 4 partitions per disk (excluding all kinds of ugly stuff like dynamic disks) or the 2TiB disk size limit
Indirectly, you won't be stuck in CSM mode (which given that it was running 7 initially, this is the case) and are able to use native UEFI mode now, and benefit from UEFI-only stuff like Fast Boot (not to be mistaken with Windows' Fast Startup) and Secure Boot

>>106531424
Sounds like a clean re-install of Windows might help here.

>LTSC LTSC LTSC
Why isn't image modification ever seriously discussed here? I know there are a bunch of removemaxxingfags around that obsess over process count like some Unix autist but that shouldn't discount the whole principle.

Also image modifications is an officially supported method for custom installs in enterprise environments, you can fairly easily do any modifications using only MS tools, though injecting GPOs is certainly harder than it should be.

There's not anything present in a default LTSC install that you couldn't do with a modified enterprise image in 10 mins. Also even regular Enterprise gets 3y security updates on any given feature release so stability isn't really a reason and MS is actually providing tools to stop install of unwanted components during feature updates like the group policy to disable particular ProvisionedAppXPackages (though this will only be released to GAC with/after 25H2).

>>106531506
These days people either change to or install Enterprise 10 and run MAS's ESU activator to keep it updated until late 2028, or install IoT LTSC 2021 to get the updates until 2032 and stay on 10.

Sadly developing open source doesn't put food on the table, nor house me (That distinction is reserved for the top) so I had to become a corporate cuck.

>>106529015
Try to keep your navel gazing to yourself.

>>106529015
>what can be done
Refuse. Refuse to participate in the use of services and products. Reuse existing hardware only.

>>106529931
>doesn't put food on the table
Help others with your skills.

>>106529931
said like a true Lockheed martin engineer

>>106529015
>its a monopoly
>dozen different companies competing with each other paying workers billions

File: linus_fanbase.jpg (79 KB, 812x621)

79 KB JPG

Aren't you supposed to be sleeping now sprout?

>>106527819
linus toilet tips?
a watercooled shitter pc.
play pc games while having a press?

>>106529065
what's the story behind that image? did he read some hardware specs on stage and the crowd went "woooo"?

File: 1747612684093111.jpg (337 KB, 1312x954)

337 KB JPG

>>106531340
I bet people give him credit for being a super genius who created Linux due to people like OP (the fags) so he doesn't just look like a guy who reads labels.

Shouldn't have laughed at Rodney Reynolds the OG YT reviewer from Leafland.

>>106531469
>Gnome
I didn't daily drive it, I added it to manjaro and it broke SDDM and the ability to get in at all! I managed to tty add and switch to GDM, which let me get into gnome and no other DE (they all said auth failure), and gnome itself was so alien I couldn't even find display settings, so I gave the fuck up on that rather than try to work with making something completely alien into what I want. So I hopped to my other root where I earlier added Endeavour (hoping my issues were Manjaro Jank but testing concluded they were not) and... never bothered to fix the Manjaro install because I use a shared /home anyways, already installed most of what I used there (likely minus some bloat) on EOS, and got comfy with console pacman. I went back to Cinnamon since it was mostly tolerable but it was eating at me whenever I saw the bad pixels, so something gave and I finally got my lazy complacent ass to try the hero of the day
>XFCE
I won't lie. I was also confused by the default UI at first, but having added xfce4 and xfce4-goodies added most of what I needed to make it my own - I figured out how to move the taskbar where I wanted, how to have it on both screens (with clocks on each, get fucked cinnamon), a familiar start menu (whisker) and my last major gripe with the UI was corrected when I discovered xfce4-docklike-plugin, added it, and then got my grouped windows with preview on hover! And 2d10h in, no opengl contexts losses, no bad pixel renders, the UI is MOSTLY as I'd like it - comfortable... I wish I had been told to go with XFCE first, because whatever everyone else is doing, it's either shit or I lost the hardware lotto, while XFCE4 Just fucking works.
What would I recommend to any new Linux user? JUST GO WITH XFCE!

most users who have spent a long time on Linux will recommend you to either go with xfce or with a bare WM setup. Linux beginners (not intended to be an insult) will prefer KDE, retards and faggots will tell you to use GNOME. KDE isn't bad per-se but it's really unstable, there's a reason *krashes* exists as a meme. there's also a reason GNOME and ebussy are heavily memed on, even more so than KDE

File: cloudflare zero trust.png (4 KB, 218x69)

4 KB PNG

>>106523316
>>106524179

NTA, but is there a way without a VPS?
I looked into something like Cloudflare Tunnel, but I wouldn't want cloudflare looking into my traffic.
I've been looking into something like it but truly e2ee, instead of just between client <-> cloudflare and cloudflare <-> server.

>>106530399
You can use ssh on termux

File: 1738253940329092.jpg (635 KB, 3840x2560)

635 KB JPG

What cheap 2.5g pci-e card for pfsense/opnsense?

>>106530966
Skip 2.5G and just get a used X520/530 10G NIC or X550 if you want RJ45 instead of SFP modules. They are dirt cheap my entire network is running on used 10G NICs, even switches and routers have come down significantly in price.

I need help planning a server. Budget is around 1.5k euros, could go higher if reasonable.
Main purpose is recording livestreams from several sites via avtdl and storing them locally, requiring a lot of storage space.
Picking hardware for it isn't the problem as it is basically just another NAS, but I'm not sure how to combine both tasks in one system.

https://developer.nvidia.com/cuda-downloads
https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

NVIDIA CUDA Toolkit 13.0 Update 1

UPDATE NOW.

Gm! Bot status? ;3

>>106529785
I started coding at 5, made my first AI startup at 16 and am already a multimillionaire and I just turned 18 yesterday. Got on that gravy train super earily.

>>106529808
>gyno genes
you mean jewish genes?

>>106530405
American tech industry runs on fake ads shown to fake users, none of this is real anon. Not even the "big ones"

>>106530423
1% of global population
60% of all billionaires

say it with me now, THERE ARE POOR JEWISH PEOPLE OUT THERE

>>106529808
thats clearly a pajeet

You're wrong :P
I told you not to use AI because it gives you bad results.

>>106531306
Not doing your homework Timmy

>on rblx
I don't want whatever's hidden in that picture

it's also okay if you don't want to :D

>>106531306
4chin strips this info or should do at least

>>106530075
it doesn't effect me. therefore, not my problem.

>>106530059
They will come for you eventually.

File: 1731742656715506.jpg (114 KB, 1280x720)

114 KB JPG

>>106530036

LOOK
AT THE PICTURE

https://www.youtube.com/watch?v=kkL72sDFYDs

>>106530036
This could be a great /x/ thread, you wasted potential posting it here. Just for the sake of clarity I'm not calling you a schizo, I'm saying /g/ will do everything but discuss this.

>>106530036
bumpinf for interest

>>106531287
I've used this for my whole life, and not about to change anything to not trigger your melties, snowflake.

>>106531331
You use a capacity tester.
https://a.aliexpress.com/_EuLYZtC

>>106531331
You empty it out, then attach a voltmeter inbetween the charger and the battery.
As (energy = voltage * current) you just measure those two and integrate over time.
Sadly both current and voltage (to a lesser extent) fluctuate a bit over the charging process. Hence >>106531348 as those do basically that, measure the power and then add it up over time.

The only reasonable way that doesn't require special equipment I can think of is charging the battery to full, then applying a constant power load (CPL). Something like a step-down module and then a resistor at its output or something. Then you can know it'll have a constant voltage at the output -> attach a resistor as a load and a LED in series, then calculate power draw of this circuit and measure the time until the LED turns off. That gives you the total energy expended without power losses from the step-down module.

something to get rid of HDCP 2.2??

>>106531348
>$30
yikes.
>>106531422
>You empty it out, then attach a voltmeter inbetween the charger and the battery.
does it work? These batteries have circuit that output constant 1.5v at all time
I think you fully charge, then use something to discharge and monitor voltage+ current until it reaches 0. But probably unreliable

>>106530066
In my experience none of the main desktops are designed for tablets. Closest ones are gnome, kde, and cinnamon. You would be better off trying to make your own based on some wlroots window manager like sway, hyprland, or niri and adding a bunch of shortcuts to the statusbar so that you dont have to use keybinds

>>106531013
>>106531042
>WPS or OnlyOffice
No libreoffice?

>>106530066
KDE is probably your best choice. The distro you want is totally up to you, but I'd recommend avoiding lts distros since you'd be stuck on older versions of KDE for years.

>>106531261
Netinstall should be good enough as long as you know what you're doing. I think it has an option to install cinnamon and/or some other desktops.

>>106531474
It's just a shittier OnlyOffice from my experience, while WPS is the most compatible with MSOffice formats.
Realistically most people use Google Docs nowadays, but for an offline office suite these two are the best.

>>106531318
also not recent

File: 1739143215048174.png (1006 KB, 644x620)

1006 KB PNG

>>106528960
>>106523317
>>106523496
>>106524181

Updated, corrected version of the previous one. As >>106524181 pointed out, all "conversations" were single turn which was an error in oversight on my end. Here's the corrected version that with actually multi turn objects within:

https://gofile.io/d/ZaBzaH

>>106524333
(I find it odd how one or two anons were hyperfixated on whether or not it had "shivers" in the dataset then the more obvious flaw of the thing only having single turn conversations. Not surprising)

>>106531349
Who do you think has the compute to finetune 2GB of data on a model large enough to be worth using?

File: 1743122148258490.gif (5 KB, 90x90)

5 KB GIF

>>106531399

1) I couldn't care less whether or not (YOU) use it, I'm just sharing it
2) Use a trainer that supports streaming.

https://docs.axolotl.ai/docs/streaming.html

You don't HAVE to load the entire thing into VRAM. Even the companies that have rooms upon rooms of GPUs don't load the entire data set into the ramp because that's an idiotic and inefficient way to do it. You little pieces of it in, train on those, then offload and reload the next piece of the data set in and train on that. Do that until you've looked over the entire data set at which point you've completed one epoch and then you do that again for the remaining epochs/steps.

You also act like training on 2 gigs of data alone is actually a lot. Sure it'll take way longer than training on something smaller like a few MB but I don't know why you think It can only be done with data center grade hardware or a giant cluster or some shit.

>>106529741
>>106530063
i was looking through some of their device specific code, from a really quick glance and knowing nothing specific about gfx906 as a target and with no concrete info it does seem to be similar kinds of optimizations in parts of libraries like composable kernel/ck-tile where AMD will use inline device assembly due to weaknesses in how HIP exposes AMD specific features
like clang only semi-recently became vaguely "aware" of buffers as like a concept and the version ROCm ships way predates that
most of AMD's really hardware specific code across ROCm has to manually pack the bitfields of a buffer into a vec4
that'd be the memory access patterns bit

also serious question does no one working on llama/ggml and/or its forks know how to use modern c++ properly or is there some kind of a project level requirement for the codebase to be like that
every time i get curious and think about trying to help i get put off by how weird the codebase is
composable kernel/ck-tile handles all that basically just using templates and constexpr, which is how you're supposed to handle that level of hardware optimization
and if you really want to go whole ass ass modern versions of clang and gcc can use constexpr std::strings as the parameters of inline assembly expressions
slam that shit on C++26 and constexpr and template as much as fucking possible wtf

We don't serve your kind here.

Guys I think this is the thread

How do I activate my RAM? Can you show me?

>>106531394
gentoo user, I actually get what I paid for ram and cpu wise everytime I update.

File: 1668940562980580.webm (31 KB, 1190x1190)

31 KB WEBM

>>106531413
The thread you say?