I feel like this really depends on what hardware you have access too. What are you interested in doing?How long are you willing to wait for it to generate, and how good do you want it to be?
You can pull off like 0.5 word per second of one of the mistral models on the CPU with 32GB of RAM. The stabediffusion image models work okay with like 8-16GB of vram.
I’d be surprised if it was significantly less. A comparable 70 billion parameter model from llama requires about 120GB to store. Supposedly the largest current chatgpt goes up to 170 billion parameters, which would take a couple hundred GB to store. There are ways to tradeoff some accuracy in order to save a bunch of space, but you’re not going to get it under tens of GB.
These models really are going through that many Gb of parameters once for every word in the output. GPUs and tensor processors are crazy fast. For comparison, think about how much data a GPU generates for 4k60 video display. Its like 1GB per second. And the recommended memory speed required to generate that image is like 400GB per second. Crazy fast.
Chatgpt is also probably around 50-100GB at most
Second this router! It had the fastest CPU and antenna vs price when I last looked. I run zerotier as a VPN on it an it works great. Plenty of ram and flash for packages too.
I think pacreport --unowned-files might be able to help with that too. Showing you files that aren’t part of any installed package. Probably only does system files though, nothing in /home
I use qdirstat a lot to determine what files are eating all my space
Your ISP knows where you’re going anyway. They don’t need DNS for that. They see all the traffic.
The symptoms you describe are exactly what happens to my machine when it runs out of memory and then starts swapping really hard. This is easy to check by seeing if disk io also spikes when it happens, and if memory usage is high
I use a pixel 2 XL, but I run lineageos 21, based on android 14. I also had the feature in lineageos 20 based on android 13.
It’s possible this is a lineageos specific feature. A quick google seems to imply that this is likely so. Unfortunate :/
Mine has a setting to not send more than one notification within X minutes I under settings > notifications > app notifications > some app > minimum time between notification sounds
Is the bad side of the seam where it stops or where it starts printing the outer wall? I assume it’s where it stops and then it cross the wall to form the infill?
To add to the PA questions, are you sure that your PA setting actually are changing anything?
What printer is this and what firmware?
Does a spiral mode print work fine?
What if you print the part significantly slower (to rule out rigidity/acceleration issues)
Instead of connecting with a web browser, can you try using curl or telnet just to check if you’re getting through at the TCP/IP connection level?
I thought I saw that Mac has the same CUPS print service/printer manager that Linux uses? In fact it seems like apple developed it. I think that helps enormously with standardizing printer configs. https://www.cups.org/doc/admin.html
I was assuming this was the government ordering the companies to. They have no incentive to do so on their own. But I believe there was a bill (which thankfully didn’t pass) that would have given the president the power to essentially order the internet turned off.
I agree that the internet is far more than facebook. But if you’re blocked at the edge of the network by your ISP, there’s really not much you can do. You’ll have access to nothing, Facebook or otherwise. Not even something low bandwidth.
If At&t, Comcast, Charter, Verizon, and T-Mobile suddenly stopped providing service to all their customers, then essentially no-one would be able to use anything on the internet at all. Even if the backbone itself (which I believe is largely owned by those same companies, but not sure) and some large datacenters that are their own isps were able to keep talking to each other, anything business or user facing would stop.
Some people who run their own mesh networks might be able to stay in contact (and people would try and start some local ones as this disaster unfolds), but that’s so few people.
How many internet service providers would have to go along before the internet was effectively off? 3? 4?
That make sense. I would use tags like that:
Flickr Published
year roundup/2022
type/Landscapes
type/Portraits
events/trips/Zion 2022
content/food
content/animals
I actually do event level as my on-disk sorting. And then tag for stuff that’s not that. But I think it would work pretty well to do the event sorting under tags as well.
Then I rate my favorite photos, usually using the green approved, not stars. But stars would work too. Then if you want to find say, favorite landscapes, the digikam interface makes it really easy to do so.
I’m not sure if you can select what tags get written into the image, but if you can, you might be able to exclude certain parts of the hierarchy, and only include content/ or type/ subhierarchies
One of the things I really like about digikam is the matching of the disk layout with the album structure. This makes it really easy to have other programs also interact with my photo library in a way that’s near impossible if you instead have an internal photo database.
Tags work great for me for multi-categorization. What feels clunky about them in your workflow? You’re even allowed to have a tag hierarchy.
For gaming like that (remote over the network), I’d recommend sunshine and moonlight. They work great if your network can handle the upload
Borg append only seems like the way to do this easily