voluntaryexilecat

still no word on how to convert/train other finetuned models into their format :(

APU board? They are going EOL soon, but these devices are built like a tank. Full Linux x86_64 support, coreboot bios. https://www.pcengines.ch/apu.htm A few sellers in the EU still have them.

Reminds me of the beginning from the novel "The Swarm" by Frank Schätzing...

I first tried it a few days ago, I'm still a bit lost. Inpainting, which is the major part of my workflow, feels not as swift as in automatic1111 and I'm still searching for the only-masked-area inpainting in ComfyUI.

But I can confirm it is much faster and uses less VRAM. And I somehow love the ability to save the entire workflow into a json. I'm missing my prompt-autocomplete plugin the most.

Do not expect you can offer this service for a competive price against cloud prices. Caring for a company IT system is a big challenge and requires more work the more users there are.

For a company this size: make a clear contract. Consider how much time you need for setup/installation, monthly hours for maintenance, monitoring and at least daily(!) backups. Let them choose if they want it with a failover and charge for the required hours and material. Also put in the contract when they can expect support from you, including a clause for a holiday substitute admin (if needed). Then put a pricetag on support hours for holding people's hands when they "can't find that file they uploaded a week ago and it is surely a server issue" and put a pricetag on engineering hours for any modifications they might want, like installing any plugins they deem useful for themselves. Hardware prices, traffic, rack space and power should be included as well. Have a good plan for updates, choose your distro wisely, do not rely on autoupdates.

Play all this through in your head, add up the hours, choose a fair rate and then you have your pricetag.

Cloud will always be cheaper, because they have their infrastructure already deployed. Building from the ground up is more expensive, but I think it is worth it. Will they?

Yes, I tested it and although it works in its current state it takes 2-3 hours per picture on Pi and 20 minutes per picture on my Desktop CPU.

But...isn't unsupervised backfeeding the same as simply overtraining the same dataset? We already know overtraining causes broken models.

Besides, the next AI models will be fed with the interactions from humans with AI, not just it's own content. ChatGPT already works like this, it learns with every interaction, every chat.

And the generative image models will be fed with AI-assisted images where humans will have fixed flaws like anatomy (the famous hands) or other glitches.

So as interesting as this is, as long as humans interact with AI the hybrid output used for training will contain enough new "input" to keep the models on track. There are already refined image generators trained with their own but human-assisted output that are better than their predecessor.

GoogleTalk once federated with XMPP/jabber, good times until their userbase was big enough to deferedate again, crippling the jabber network. It will happen again if we let it.

Metas plan is to draw users into their network and use the fediverse as an initial catalyst ("look! so much content already there!"). Once their userbase is large enough, they will deferate again claiming protocol difficulties or something equally vague, but they will just want to start rolling out advertising which would not be displayed to users from other instances. Most users will not keep two accounts and jusy stay with the big corp and leave the original fediverse again.

Yes, that should work. Check out stable-diffusion-webui (automatic1111) and text-generation-webui (oobabooga). And grab the models from civitai (stable diffusion) and huggingface (llms like llama, vicuna, gpt-j, wizard, etc.).

Check out Stable Diffusion and the llama model family. You can run those offline on your local hardware and wont have to worry about sharing private details with some cloud service that openly says they will look at your discussions and data and use it for training.

The short tinnitus that lasts just a few minutes is relatively common. Most common cause is stress and circulation issues. There seems to be no alternative name for the short tinnitus to differentiate between the permanent ringing.

I found that if it starts ringing in my ear due to stress or just spacing out during overthinking stuff, hyperventilating (increase blood oxygen levels) briefly and massaging my ear canal (increase circulation) from the outside helps to get rid of it more quickly. Maybe this helps somebody someday.

For composition I use Semantic Segmentation ControlNet - Sketch loosely, then inpaint or more ControlNet. Of course I use GIMP or any other tool to finetune the image or to "force the model's hand" a little during inpainting.