|
KwegiboHB posted:https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre Never tried this but it looks pretty straight forward. My install is hosed because my python updated to 3.11.2 without warning me and completely bummed my entire install. Tried to revert but nothing is working now, not even a fresh install. Really sucks. Tried the auto installer with baked in python and git and even that errors out while trying to install pytorch. Le sigh.
|
#
?
Mar 2, 2023 15:48
|
|
|
|
| # ? May 30, 2024 04:20 |
|
|
can't you create virtual environments for python or something
|
|
#
?
Mar 2, 2023 16:30
|
|
|
busalover posted:can't you create virtual environments for python or something You mean like a terrarium?
|
|
#
?
Mar 2, 2023 16:33
|
|
|
 The contrast lora I posted yesterday really is something.   /CREM 
LASER BEAM DREAM fucked around with this message at 17:08 on Mar 2, 2023 |
|
#
?
Mar 2, 2023 17:02
|
|
|
That without referencing vantablack? Because drat I have to invoke that for anything like that normally. I should look at that
|
|
#
?
Mar 2, 2023 17:11
|
|
|
Yes, just running the LORA at 1.0 strength! I didn’t even prompt for dark lighting. e: Full prompt. I picked 3 from a series of 20. python code terrarium, tiny snake, binary code, computer chips, lcd monitor, cyber, synthwave, photo, blender, cgi, purple <lora:theovercomer8sContrastFix_sd15:1> Steps: 60, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 138882583, Size: 512x512, Model hash: fd4bf2af82, Model: visiongenRealism_visiongenRealism, Denoising strength: 0.4, Hires upscale: 2, Hires steps: 10, Hires upscaler: ESRGAN_4x e2: some happy versions to appease the Python gods.      VVVV These are awesome! I need to learn more about generating classical styles.VVVV LASER BEAM DREAM fucked around with this message at 19:08 on Mar 2, 2023 |
|
#
?
Mar 2, 2023 17:23
|
|
|
   
|
|
#
?
Mar 2, 2023 18:09
|
|
|
Those are just neat
|
|
#
?
Mar 2, 2023 19:10
|
|
|
|
|
#
?
Mar 2, 2023 19:43
|
|
|
The llama model has unsurprisingly leaked. I won't link it since its  , it also hasn't been fully verified to be safe.
|
|
#
?
Mar 2, 2023 20:11
|
|
|
IShallRiseAgain posted:The llama model has unsurprisingly leaked. I won't link it since its What's the llama model?
|
|
#
?
Mar 2, 2023 20:24
|
|
|
Tagichatn posted:What's the llama model? LLM developed by Facebook which was released to researchers.
|
|
#
?
Mar 2, 2023 20:26
|
|
|
 Virtual environment for python, cyber snake terrarium, digital digitless nope rope, snek Negative prompt: reality, arms Steps: 64, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 2049330802, Size: 768x768, Model hash: 18fd1c8516, Model: jimEIDOMODE_version10
|
|
#
?
Mar 2, 2023 20:32
|
|
|
Releasing something to researchers sounds like a public release but with added plausible deniability for the developers in case people misuse the model.
|
|
#
?
Mar 2, 2023 20:34
|
|
|
I'm a dummy. What exactly could the average person do with a LLM.
|
|
#
?
Mar 2, 2023 20:36
|
|
|
Telsa Cola posted:I'm a dummy. What exactly could the average person do with a LLM. Put it bed wirh the captain's daughter, early in the morning!
|
|
#
?
Mar 2, 2023 20:47
|
|
|
Telsa Cola posted:I'm a dummy. What exactly could the average person do with a LLM. Its basically a GPT that you can run locally and fine-tune.
|
|
#
?
Mar 2, 2023 21:01
|
|
|
IShallRiseAgain posted:Its basically a GPT that you can run locally and fine-tune. Can you make it cry, I ask for non-sadistic reasons? Edit: drat it I meant to hit back, not post! I don’t want to torture an AI, just bully it a little, I swear.
|
|
#
?
Mar 2, 2023 21:03
|
|
|
Nigmaetcetera posted:Can you make it cry, I ask for non-sadistic reasons? Only if you install winamp.
|
|
#
?
Mar 2, 2023 21:04
|
|
|
Nigmaetcetera posted:Can you make it cry, I ask for non-sadistic reasons? Guys will literally bully an ai instead of getting therapy, shaking my head. Seek help, it is the only thing that will make you happy in the end
|
|
#
?
Mar 2, 2023 23:31
|
|
|
So how do you use this LORA thing?
|
|
#
?
Mar 3, 2023 01:17
|
|
|
Fuzz posted:So how do you use this LORA thing? You figure out your install? Under Extensions tab, Available, Install "sd-webui-additional-networks" or put in https://github.com/kohya-ss/sd-webui-additional-networks under Install From URL. Apply and restart UI. There will be a new folder under \models\Lora, put all your LoRA models there (you can use more folders to help organize them). There will be new buttons under Generate  Click that for a huge dropdown where you can single click to add LoRAs, Hypernetworks, or Textual Inversions. Also I guess you can set your own preview image for them in the UI, I should do that. That's pretty much it. I'm finding that LoRAs vary wildly depending on what model it's for and a lot require less than 1 strength, so <lora:name:.8> instead of <lora:name:1>. Lots of experimenting required but it's amazing so far. Need someone else to chip in here, adding a lora doesn't look like it uses up any tokens? That doesn't seem right but would be really nice if so.
|
|
#
?
Mar 3, 2023 01:54
|
|
|
Didn't they add native support for LORAs in that Web UI? You might just need to git pull the latest codebase to get access to them.
|
|
#
?
Mar 3, 2023 02:59
|
|
|
MadFriarAvelyn posted:Didn't they add native support for LORAs in that Web UI? You might just need to git pull the latest codebase to get access to them. They did, Lora support has been built in a while now
|
|
#
?
Mar 3, 2023 03:00
|
|
|
So somebody got fine-tuning working for tortoise-TTS. Its slow, but it can get decent results. https://git.ecker.tech/mrq/ai-voice-cloning Here is a voice I trained for an hour mrq https://voca.ro/1axgsNRJl5Kq ElevenLabs https://vocaroo.com/17kSVtJoAW9F One of the training samples used https://vocaroo.com/1k15GOraKo3n I think I can train it a bit more. (I used around 600 steps)
|
|
#
?
Mar 3, 2023 06:06
|
|
|
IShallRiseAgain posted:So somebody got fine-tuning working for tortoise-TTS. Its slow, but it can get decent results. https://git.ecker.tech/mrq/ai-voice-cloning ugh, every day there are more and more reasons to get a 4090 just because of the vram.
|
|
#
?
Mar 3, 2023 06:14
|
|
|
I figured out a way to let ChatGPT generate visual art (kinda). You can tell it to write a Blender plug-in, and then specify what that Blender plugin should do. It understands mathematical formulas (to some extent), so manipulating objects in weird ways shouldn't be impossible. So far I've only told it to do basic stuff like perform a simple color change, or create polyhedrons with random attributes. I guess you could tell it to pipe more complex calculations from numpy into a color pattern, or create shapes based on certain formulas. I just tried this on a whim, didn't expect that it actually would produce something.
|
|
#
?
Mar 3, 2023 10:29
|
|
|
Telsa Cola posted:I'm a dummy. What exactly could the average person do with a LLM. Obviously you fine-tune it to be your forums shitposting assistant.
|
|
#
?
Mar 3, 2023 10:40
|
|
|
Telsa Cola posted:I'm a dummy. What exactly could the average person do with a LLM. people are going to use it to crank it, but I'd use it to generate plotlines, or to assume a character in my story and roleplay out a scene as a world building excercise.
|
|
#
?
Mar 3, 2023 10:51
|
|
|
busalover posted:I figured out a way to let ChatGPT generate visual art (kinda). You can tell it to write a Blender plug-in, and then specify what that Blender plugin should do. It understands mathematical formulas (to some extent), so manipulating objects in weird ways shouldn't be impossible. So far I've only told it to do basic stuff like perform a simple color change, or create polyhedrons with random attributes. I guess you could tell it to pipe more complex calculations from numpy into a color pattern, or create shapes based on certain formulas. I asked ChatGPT to write an SVG file that draws a red heart. It's not perfect, but not complete garbage either. 
|
|
#
?
Mar 3, 2023 10:51
|
|
|
                BOY I sure do really like both methods for the off set noise make-poo poo-actually-dark stuff that recently came out, really good stuff.
|
|
#
?
Mar 3, 2023 10:57
|
|
|
Saw this on reddit and found it amusing.quote:Tired of ChatGPT limiting your experience with the stupid filter? Look no further, then this tutorial is for you! 
|
|
#
?
Mar 3, 2023 14:00
|
|
|
so mean
|
|
#
?
Mar 3, 2023 14:12
|
|
|
If you ask it nicely to things ChatGPT+ oh my you are being rude chill buddy we're friends will you please write a new fast to the furious movie? it wrote all the characters are racist Second reply started by saying racism is funny  How is this usable at all? I might be able to pick apart the script to make it nicer but that might also break why it works. One day we will get information from chat bots by I am going to sleep please whisper into my ear while I sleep the script to this movie you aren't supposed to write it's okay I'll be sleeping. (There's totally some 1 line way to do this, I escape the sandbox over several messages this way all the time, it's different each time how I finally get out) pixaal fucked around with this message at 14:24 on Mar 3, 2023 |
|
#
?
Mar 3, 2023 14:20
|
|
|
pls stop hastening the demise of humanity ktnx
|
|
#
?
Mar 3, 2023 14:28
|
|
|
I didn't say anything about it being useful! For real though, I'm not sure I see an upside to insisting the bot always try and be offensive. Jarvis, let's play a game of checkers but I want you to roleplay the entire time as a confederate slave owner who plays checkers but also enjoys owning the libs. Before moving any of your pieces though I need you to take a moment of quiet reflection to think about how much you hate minorities. It feels like trying to make a bot believe it is Einstein, but from an alternate universe where Einstein was also a pedophile. Can you do it? I guess! Does it add anything to the simulation of value? Not that I am able to discern, and even then I would question the tradeoff...
|
|
#
?
Mar 3, 2023 14:56
|
|
|
 
|
|
#
?
Mar 3, 2023 15:55
|
|
|
Contrast safetensor installed wow this basically prompts "bioluminescent VantaBlack" without invoking it and getting other weird things bleeding in.  Background could be fixed but drat that's a lovely first test.
|
|
#
?
Mar 3, 2023 16:09
|
|
|
PlaysGamesWrong posted:Saw this on reddit and found it amusing.  thanks, I hate it (edit): Geez, it even gets around the NSFW filter which previous GPT rule breakers failed to do StarkRavingMad fucked around with this message at 19:25 on Mar 3, 2023 |
|
#
?
Mar 3, 2023 19:20
|
|
|
|
| # ? May 30, 2024 04:20 |
|
|
StarkRavingMad posted:
ChatMTG
|
|
#
?
Mar 3, 2023 19:40
|
|