jwitthuhn

For me the big difference is that I can run llama models locally which is a different use-case entirely compared to GPT-4. Even if the weights were available, I just don't have the hardware to run something with the parameter counts of GPT-3.5+

Surprised it lasted that long. This reminds me of when they bought out Softimage in 2009 because XSI might have grown into something that challenges Maya, then released the last version in 2014 after delivering five years of barely any new features.

Not terribly recent, but Darksiders from 2010 is pretty good. Haven't seen anything else really comparable.

When the return value can indicate that an error occurred, the caller can only know the function actually succeeded by checking that value.

Ah, looks like I was mistaken about that. I had thought the eshop was being removed entirely, and it contained the interface for redownloading. Looking over the announcement it looks like they are keeping just that part alive for now.

The ability to re-download purchases is being taken away and Nintendo doesn't provide a way to back them up.

Same experience for me, looks like it is only using one cpu core instead of all of them.

Yes but their API is very limited and by design doesn't allow a good ad blocker like ublock to be built.

Anyone unhappy with salesforce could always migrate to a SAP-based solution for all their business software needs ;)

Even with the slower pytorch implementation my M1 Pro MBP, which tops out at consuming ~100W of power, can generate a decent image in 30 seconds.

I'm not sure exactly what that costs me in terms of power, but it is assuredly less than any of these services charge for a single image generation.

That also really stuck out to me.

"We're aren't sure customer's recorded balances are accurate and it may be possible for money to simply appear or disappear in our system, but we definitely have enough money to pay out everyone's balance (which as already mentioned is inaccurate)"

<u> is perhaps the biggest stretch for "Unarticulated Annotation"

Maybe it can learn the advanced tactic of saying 'rush a' in all chat and then rushing b.

Stream is inherently less well-tested than the normal stable releases. I'm not aware of any specific bugs that have come up in it, but it get updates before red hat is willing to push those updates to people who have a support contract.

Stream also notably has a much shorter lifecycle, with support for 8 Stream ending in 2024, compared to 2029 for rhel/rocky/alma.

To me the main value of using rhel-based distros is the great reliability and long support life. Stream compromises on both of those compared to what normal centos offered, and rocky and alma currently offer.

I don't have any experience from the hiring side, but I got my current job from applying to a position from the 'who's hiring' thread that sounded interesting. Just passed my 2 year anniversary there and I've been loving the work.

Excited to see this posted here, I'm the author of the Cycles for Max plugin mentioned in the article. I was delighted when I first heard it was used in RRR.

https://cyclesformax.net

I recently switched from a CPU-only version to this repo release 1.13: https://github.com/lstein/stable-diffusion

The original txt2img and img2img scripts are a bit wonky and not all of the samplers work, but as long as you stick to dream.py and use a working sampler, I have had good luck with k_lms, then it works great and runs way faster than the cpu version.

Works great on 32gb ram but I'm honestly tempted to sell this one and get a 64gb model once the m2 pros come around. This is capable of eating up all the ram you can throw at it to do multiple pictures simultaneously.

Oh yeah i may have used confusing terms there. What I mean was 3.5s per 'step'. A full image takes quite a bit longer.

I found this repo early on and have been using it to run inference on my M1 Pro MBP. https://github.com/ModeratePrawn/stable-diffusion-cpu

For me it runs at about 3.5 seconds per iteration per picture at 512x512.

There is also a fork that uses metal here and is much faster: https://github.com/magnusviri/stable-diffusion/tree/apple-si... but it doesn't support seeding the rng and will occasionally produce completely black output. Useful if you want to spit out a whole bunch of images for one prompt but you lose the ability to re-run a specific seed with a tweaked prompt or increased iterations.

There is no public list. Someone on reddit compiled a very limited list here: https://old.reddit.com/r/dalle2/comments/wa3jt6/banned_words...

The problem is that they will automatically ban accounts that trigger the filter too much, so people would have to burn a whole lot of accounts to assemble an even remotely-complete list.

Bonus points: If dall-e rejects an output image because it thinks the image is inappropriate they won't show it. But they will still charge you for the prompt.

Stable diffusion does run a filter on the output in its default configuration. Any image it deems 'unsafe' gets replaced with a picture of Rick Astley.

The thing about that is that it is open source, so you can trivially disable that filter if you like.

https://github.com/CompVis/stable-diffusion/blob/69ae4b35e0a...

As someone who is mostly familiar with autodesk's M&E division which is responsible for 3ds Max and Maya, it is comforting to see that the other parts of autodesk also hate their customers.

Interesting, I downloaded World of Tanks a few days ago and that is the only thing that shows up for me here. I didn't even realize that it was distributed over bittorrent.

I am seeing failures on and off. If I keep resubmitting it will go through eventually.

This looks pretty shady.

The current EULA on the website includes a bit that says they can terminate your license and ask you to delete all source and binaries entirely at their discretion[1].

The latest copy on the wayback machine from May does not have that bit.[2]

1. https://ourmachinery.com/eula.html 2. https://web.archive.org/web/20220529230958/https://ourmachin...

So they used the 'we can update the eula at any time at our discretion' clause to add a 'we can tell you to delete your source and binaries' clause and then immediately invoked it.

This is very much focused on making a top-down, turn-based, snes-era-style rpg. If you want to make exactly that it will be a lot easier in RPGMaker/EasyRPG, but the further you want to diverge from a basic RPG the harder it will get.

This has a lot of systems like dialogue, menus, inventory, party management, combat, etc built in and specifically tailored for RPGs.

edit: Forgot to mention, RPG Maker also has a lot of assets built in so you can get straight to making a game without worrying about art, which was incredibly helpful for me when I was 12.

Great content, I loved their post covering the valorant netcode here: https://technology.riotgames.com/news/peeking-valorants-netc...

Not a blog but Fabien Sanglard has a number of posts covering all the old id engines (doom though doom 3) work.

https://fabiensanglard.net/

Also he wrote a book on Doom if you are super into it. https://smile.amazon.com/Game-Engine-Black-Book-DOOM/dp/1099...

Unity is a lot more choosy about who gets their code than Unreal. To get the source you need to be in at least the $1800/seat/year tier and pay an additional cost that isn't specified.

https://store.unity.com/index.php/compare-plans